Hmmmmm...I prefer LESS stocks rather than MORE in my real life portfolio...maybe it's time to cut some of the laggards in my CAPS port. Interesting pattern so far though...does anyone recall if it's always been the players with less picks in the top? I just now started to pay attention, so I don't know what the history is.
Hey Bruce,I don't think so...I've seen Seth Jayson and Bill Mann and Charly Travers all rated highly at various points with lots of picks. Of note, look how many Seth has ended. Best,bd(who is sporting an all new and exciting CAPS port style!)
Interesting pattern so far though...does anyone recall if it's always been the players with less picks in the top?My gut feeling is that it's temporary and due only to the influx of thousands of new players into the game all at once. I think over a longer period of time that in general small pick lists will not be able to generate the magnitude of total score needed to stay in the top ranks. Now what size pick list is "optimal" I have no idea.
I think over a longer period of time that in general small pick lists will not be able to generate the magnitude of total score needed to stay in the top ranks.Charly,I don't know. I mean, we're seeing top tens with less than 50 points. And that's compared to your 600 and Russell's 900? That just doesn't make sense to me. I really think accuracy is weighted too heavily. What is long-term, and does that mean that someone is going to have to have 10,000 points to overtake someone with high accuracy and only 100 points?If I were paying someone to invest for me, I'd have to go with Russell (okay, maybe you), simply because he's outperforming the most. I don't care one iota if he's the most accurate. I want the most money, period.I understand that in order to keep the game interesting to new players, there needs to be an accuracy benefit...but again, I think it's too much.Just my thoughts.Ryan
I don't know. I mean, we're seeing top tens with less than 50 points. And that's compared to your 600 and Russell's 900? That just doesn't make sense to me. I really think accuracy is weighted too heavily. What is long-term, and does that mean that someone is going to have to have 10,000 points to overtake someone with high accuracy and only 100 points?Take a look at the players list sorted by score, and you will see that the trip down from Eldrehad's 850 to TrackFortune2006's 96 points is pretty quick, only 23 players in between those two (by the way, don't forget to add me as a favaorite as you make your way down!) Couple that fact with the fact that score and accuracy are ranked in percentile terms and it quickly becomes clear why new players with very few picks occupy the top spots currently. Heck, even a score of 20 gets you into the 80th percentile right now!Now imagine, if you will, the situation a few years rom now, when a subset of the 15,000 new players will have had 2 years or more to build up their scores into the 1000-5000 point range. Even if only 10% of those 15000 players will have made the trek up to a 1000 score, the roster will be full enough that new players who strike it lucky with 100% or 85% accuracy from the get-go will not get their score high enough to send them over to all-star land.Give it time, this will all work itself out.-ER
My gut feeling is that it's temporary and due only to the influx of thousands of new players into the game all at once. I think over a longer period of time that in general small pick lists will not be able to generate the magnitude of total score needed to stay in the top ranks. Now what size pick list is "optimal" I have no idea.I'm personally convinced that this isn't a temporary phenomenon, and that big pick lists have zero chance of cracking the top spots in CAPS.The problem is accuracy.In order to get in the very top spots, one has to do very well in both score and accuracy. Well, accomplishing the latter with a big pick list is darn near impossible.Take your portfolio for example. As I write this, you have an accuracy of 69.35%, or 86/124 (adjusting for ended picks with scores between 0 and 5 which don't count).The odds of doing this well or better by randomly selecting 124 tickers, each with a 50% probability of being accurate, is about 1 in 103,000. A new CAPS player, however, by randomly selecting 7 tickers, will beat your incredible 1 in 103,000 feat between 22% and 23% of the time.Big pick lists have zero chance of cracking very top spots in CAPS.To put it another way, the smallest accuracy in the current top 20 is 86.36%. In order to beat this accuracy by adding new picks, you'd have to add 155 accurate picks to catch up (making you 241 for 279). The odds of pulling thiss off by randomly selecting tickers is about 1 in 4.5 x 10^46. That's a 4 followed by 46 zeros!The lowest accuracy in the top 60 is 75%. To beat this you'd have to add 28 accurate picks. This is much easier than our example above. This one is only about a 1 in 270 million shot.While there have been times in the past in CAPS where big pick lists like yours, and mine, have been able to get into the top spots in CAPS, that was when there were only a few hundred players. With thousands, and with it so easy for so many of those thousands to clobber our accuracies by doing nothing more than randomly selecting tickers, I think those days are long gone.Now, as CAPS matures, is it possible that some of these players with small pick lists might make more picks in an effort to boost score, and widening the distributions of accuracies a little bit? You bet. But they'll soon find themselves in the same dilemma you and I are in, namely that the score that they get isn't worth the accuracy that they are giving up.Let me put this one final way... let's use the most extreme case of all as an example - let's use mine.I am the top scorer, as of this writing, in all of CAPS.As I write this, the person sitting in the #10 slot is moobucks. moobucks' accuracy is 92.31% over 13 picks (12/13) - an impressive feat. Going 12 for 13 or better by randomly selecting tickers is about a 1 in 600 shot.moobucks' score is 42.74 while mine should be around 950 (there's a reverse split problem on FSYS that needs to be corrected in CAPS which is currently costing me 100 score points that it shouldn't be). My accuracy is (adjusting for ended picks with scores between 0 and 5) 65.52% or 228/348.The odds of going 228 for 348 or better given a 50% probability of success for each pick are, hold onto your hats folks, 1 in 270 million.So, I have the highest score in all of CAPS (granted, helped by my big pick list size) which is more than 22 times greater than the person sitting in the #10 slot, have pulled-off a 1 in 270 million shot in accuracy, which is about 460,000 times harder than the #10 player's 1 in 600 shot, and am 254 places behind this player in the overall rankings.Again, big pick lists just have no chance.Now, big pick lists sure are a lot of fun - and I certainly learn a lot by looking at different stocks - and I love the 'kid in a candy store' feel I get when making new picks.But if I had to do this all over again, and my primary goal were to make it to the Top Fool spot - knowing what I know now, I'd make the most judicious 7 picks I could, never ever make a change to that pick list, and hope for the best. The more picks one makes, the more one is likely to slide down the accuracy scale - and I think that the size of the pick lists currently in the top spots as an example, plus the math I've laid out here, make a pretty compelling argument that when it comes to getting into the Top Fool club, a small portfolio is the way to go.Regards,Russella.k.a. TMFEldrehad
I am going to disagree with you in part. Small pick lists with high accuracy have a short term advantage only. Their total scores will be passed in a matter of months by thousands of players with mediocre accuracy. A 99th percentile accuracy rating due to a small pick list and good fortune on those picks won't be so powerful an influence when that player's total score is in something like the 30th percentile, or worse, for example. This is where ultimately I expect most of these players to end up over time barring one of their 8 picks being a 5 bagger or something like that.I do agree that big pick lists will struggle if they cannot maintain good accuracy. Most people with a lot of picks will probably be between 50-60% correct and that's ok but not good enough to get to the top by any stretch.Small pick lists also have concentration risk. Say a person is in the green on 7 out of 8 picks. Then one pick goes bad and suddenly the sterling 87.5% accuracy plummets to 75% and the player rank craters with it. Someone with 40-50 picks won't even blink if one pick goes bad.Total score is the achilles heel of the small pick list. The CAPS beta hasn't been open long enough to expose it.I think the winning formula will be a "medium" size pick list with very good accuracy. This can generate both total scores and accuracy above the 90th percentile.
Charly --Small pick lists with high accuracy have a short term advantage only. Their total scores will be passed in a matter of months by thousands of players with mediocre accuracy.Do you think there will be that many players with high point scores? Fools know their stocks, but don't most individuals tend to equal or do worse than the S&P 500? I predict after 2 years and 10,000+ players, most will have negative scores-- so anyone with 100% positive accuracy should have an unusauly high points percentile as well.
I am going to disagree with you in part. Small pick lists with high accuracy have a short term advantage only. Their total scores will be passed in a matter of months by thousands of players with mediocre accuracy.For the average small pick list with a high accuracy? You may very well be right. For the successful small pick list with a high accuracy? Not a chance in a million that a big pick list (on the order of yours, or worse, mine) will catch them.My case is a great illustration because I am currently #1 in score, but you are pretty much in the same boat, being #3 in score. Since my case is the easier with which to demonstrate where I'm going, I'll use mine (again, though, you, Otter, Bent - all in the same boat).Let's see what I have to do to crack the current top 10. The rankings have changed a little bit since my last post, so to crack the top 10 as of this writing, I have to unseat jburger124.jburger124 has a current score of 43.90 and a current accuracy of 87.50% (or 7/8). In terms of score, jburger124 is rankend #107, and in terms of accuracy, this player is tied for #110.So, being #1 in score that means I am 106 places ahead of jburger124 in that category. Since score counts twice accuracy, that means if I am 212 places behind jburger in accuracy, we'll be tied overall in player ranking - to unseat this player then, I have to be no worse than 211 places behind this player.Well, that means to crack the top 10, I need to be no worse than #321 in accuracy, while maintaining my #1 spot in score.Well, #321 in accuracy is currently held by someone with an accuracy of 85.71%. Actually, it's a big tie at #158 with a whole bunch of players who have accuracies of 85.71%. The next spot after the tie is #379 with an accuracy of 85.29%.Well, first point is that I actually have to improve to #158 or better in accuracy, isn't it? I can't actually get to #321, it's either #379 or #158 - there is no in between.And this is a big point! I wrote something on the 'power of simple fractions' some time ago, and this is another illustration of it. Accuracies are not even remotely close to being normally distributed, and that is a big part of the problem I see with the disadvantage of big pick lists.But let's be generous and put the power of simple fractions aside, and say that instead I'll crack the top 10 if I'm only #379 in accuracy (at 85.29%) instead of having to be tied for #158 (which is actually the reality of my current situation).Well, again, I need an accuracy of 85.29% or better just to crack the top 10, and this assumes that I remain #1 in score.This, my friend, is impossible.Again, I am currently 228/348. To get an accuracy of 85.29% in my current situation, I would need to add 467 successful picks to my portfolio, without one single inaccurate pick along the way.The odds of this?Gee, only about 1 in 381,072,821,083,495,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 Forget the 'average' small pick list for a moment, to crack the top spots, we big pick listers have to beet the 'well above average' small pick lists.And again, we have no chance.Want to know what's worse? As startling as the math above is, it is getting worse and worse by the day. I am #1 in score, can't move up there - granted, new players aren't likely to push me down either - but new players are highly likely to push me down in accuracy - again, about 22-23% of the time by randomly guessing alone - and increasing my percentile ranking in accuracy is the only opportunity I have to catch those currently ahead of me. The more players come to CAPS, the harder it gets (as if the above case didn't spell out it's impossible as it is).And again, although I'm using myself as an example, this isn't about me and how hard it would be to get to a top spot in CAPS - it's about big pick lists and the fact that I am completely convinced that given the way CAPS is currently structured, they have no chance.Don't get me wrong, I'm not saying that the leader board will eventually only have players with 7 picks... but it will always and forever be dominated by players with smaller pick lists. How small? I'm not certain, but my guess is no more than 20 or so, but no question not in the hundreds, which is where you and I and Otter and Bent find ourselves.Regards,Russella.k.a. TMFEldrehad
Russell,My inclination is to agree with Charly, that a mid-size pick list with high accuracy should come to dominate. At least among the all-stars and higher, the top 10 may well always be dominated by small accurate lists containing a home run+. I had a brief sip of coffee with the top-20 late last week when one of my selections fluttered to positive. It made me hesitant to add new selections since I didn't want to risk the accuracy. I went ahead and added selections. What the heck, I'm here to see how well I do with my picks and learn from mistakes, not so much to strive for top positions.What kills me is a short I want to end that sits just under +5%. Bounced to +5.01% today, but I was afraid to end it and find that I just missed due to price delays (burned you once or twice as I recall). I really want to lock in the permanent accuracy rating on that one.Someone should set that reclusive Russian mathematician onto the problem.Zz
Just to follow-up:I think the winning formula will be a "medium" size pick list with very good accuracy. This can generate both total scores and accuracy above the 90th percentile.This actually proves my point in a way.There is no way possible for a even a "medium" pick list to achieve the 90th percentile in terms of accuracy. Not when, among those randomly selecting 7 tickers, the 77th percentile (let alone the 90th) clocks in at an accuracy of 5/7, or 71.43%.In other words, about 23% of players randomly choosing 7 tickers will have accuracies of 71.43% or better by virtue of sheer luck alone.The odds of a 'medium' pick list, of say, 30 tickers even getting to the 77th percentile (let alone the 90th)? About 1 in 124.Again, that's only to get to the 77th percentile (when compared to people who only make 7 picks).Let's put this another way. about 6% of all players choosing 7 stocks will get accuracies of 85.71% or better. Our 'medium' pick list player with 30 picks has a 1 in 34,000 shot of catching this player in accuracy. So, our big pick list, just to stay even, has to pull off a feat that is about 2,000 times harder.I'm sure some Fool is thinking, "But what about score? It counts twice as much as accuracy!"What about it? All things being equal, our medium pick list will have a score about 4.3 times higher than our small pick list player (with 7 picks). So, our medium pick list gets 4.3X the score (and since score counts twice accuracy, 8.6X the score 'boost' in the rankings), and all he has to do to get that is make it 2,000 times harder on himself just to tread water on accuracy.No, the math doesn't work quite this way, as I'm not taking into account distribution of score results and the 8.6X isn't really a 'true' comparison, but I think it does beautifully illustrate the point, even if the math isn't pure.Here, perhaps, is what I think is the biggest point of all:Assuming one starts out with 7 picks, increases to score by adding each successive new pick is a linear function. Furthermore, increases to player rank coming from are also linear, even though they get the 2X boost from score counting twice accuracy, it's still linear, just with a different slope - the increased difficulty in maintaining one's accuracy as one makes new picks is, however, is an exponential function.And therein lies the crux of this problem. It's only a matter of time until the exponential 'handicap' begins to outweigh the linear 'benefit'.And while I can't say for sure, I think that crossover point is, at most, at around 12-15 picks. Clearly, it is well before 100 picks - again, making it nearly impossible for a big pick list to crack the leader board.Regards,Russella.k.a. TMFEldrehad
Heh, the main point, I took the touble to bold, is the most horribly written paragraph in the post...Let me try again:Assuming one starts out with 7 picks, increases to score by adding each successive new pick is a linear function. Furthermore, increases to player rank coming from score are also linear, even though they get the 2X boost from score counting twice accuracy, it's still linear, just with a different slope - the increased difficulty in maintaining one's accuracy as one makes new picks is, however, an exponential function.
Interesting thoughts....my brain hurts too much to process it all right now, but FYI it struck me that the description of the "small pick list of prodigiously successful stokcs" while "never changing them" reminds me a heck of a lot of Warren Buffet. Regards,Doc
Let's put this another way. about 6% of all players choosing 7 stocks will get accuracies of 85.71% or better.And of this select group, what percentage are likely to have a total score that is decent enough to capitalize upon their accuracy and put them on the leader board? Not based on today's scores, because CAPS hasn't been running long enough, but based on possible expected returns over the next year for example.That requires some fancy modeling I think. If it is indeed highly likely that the CAPS top ten could be entire populated by random chance, then I don't think that's good for the game. Ideally the top ten players would be good stock pickers. If this is a problem, some solutions could be nudging the minimum pick requirement for a rating from seven up to ten and/or requiring a bare minimum of making 5 new picks per rolling 12 month period to maintain some base activity level.
And of this select group, what percentage are likely to have a total score that is decent enough to capitalize upon their accuracy and put them on the leader board? Not based on today's scores, because CAPS hasn't been running long enough, but based on possible expected returns over the next year for example.Here's my point - it doesn't matter how many do, just that 10 do. If 10 do, they will dominate the leader board and make it impossible for either you, or I, to ever catch them.Let's put this another way, let's assume that although their accuracies are randomly distributed over their 7 picks, none of them ever get a positive score, and that every single one of these new players has a score of -10,000.About 23% of them will still have accuracies of 71.43% or better. About 23% of all new players, even with scores that are horrible, will push both you and I further down in player ranking because they are 'stepping between' us and the people who are currently on the leader board (in terms of accuracy) - making it that much more difficult for we big pick list folks to catch them.Again, what I'm saying is... it doesn't matter. We can't catch the people who are on the leader board now, and more and more players, even if they all have horrible scores, will only make it harder. Furthermore, there will, hopefully, always be new players - perhaps not the influx we've seen over the past week, but there will still always be new folks joining CAPS - and about 23% of them will push the big pick lists down in the rankings relative to the small pick lists.If it is indeed highly likely that the CAPS top ten could be entire populated by random chance, then I don't think that's good for the game.I am only using random examples to illustrate the point. No, I don't think the top 10 should be people who got there by random chance, nor do I think it ever will be. What the 'random' examples help to illustrate, at least in my mind, is much more difficult it is for big pick lists to get to the leader board than it is for small ones. The top 10 won't be people who chose randomly, but the top 10 will be those with smaller pick lists who chose well.If this is a problem, some solutions could be nudging the minimum pick requirement for a rating from seven up to ten and/or requiring a bare minimum of making 5 new picks per rolling 12 month period to maintain some base activity level.I've been advocating increasing the minumum pick list size for some time now for exactly these reasons (though I'd personally go with at least 20). I do not like the idea of forcing players to make new picks as time progresses, however. The reason is simple. If big pick lists are at a disadvantage, those who've been playing CAPS the longest, under this rule, would be at the biggest disadvantage of all as they'd be forced to have the biggest pick lists.Regards,Russell
Here, perhaps, is what I think is the biggest point of all:Assuming one starts out with 7 picks, increases to score by adding each successive new pick is a linear function. Furthermore, increases to player rank coming from are also linear, even though they get the 2X boost from score counting twice accuracy, it's still linear, just with a different slope - the increased difficulty in maintaining one's accuracy as one makes new picks is, however, is an exponential function.Good argument.So a players rating is somewhat analogous to stock valuation using PEG? Bill Mann's not gonna like this.Yet, the accuracy number does enable new players to experience some level of success right off the bat. This will likely be crucial to the long term appeal of the game / project. Maintaining that rating over time will be very difficult, especially if one regularly adds new picks.The poster that pointed out that Warren Buffet would likely do very well in CAPS is on to something methinks. Choose wisely, choose slowly, focus more on not loosing than winning, lock in gains and accuracy where and when appropriate.ZzP.S. - Where are the new players complaining about the head start in score that alpha testers got?P.P.S. - Before going 'live' it should be suggested that existing 'test' players be given the option of starting fresh.
It is also interesting to me that currently a 50% accuracy, which statistically should always be ~ the 50th percentile, is currently only good for the 39th percentile.Perhaps the Fool has been quite successful in it's mission to Educate and Enrich? - And yeah, they're still funny too.It will be curious to see if the 50% accuracy rating reverts to the mean over time or as the game is expanded ever outwards.Zz
Alternate scoring solution--If the player rating were determined by taking your SCORE (not your score PERCENTILE) multiplied by 2/3 of YOUR OWN accuracy (not your percetile) -- then ranking all players on that basis by percentile, wouldn't it make more sense and solve all of these problems?So under my scoring method -- take 4 players and their scoresName------------ Score Accur. S*2/3(A)TMFBreakercharly 680 67.75 30713Elderhad-------- 860 66.09 37891padraig23**----- 60 100 4000Se7ensparks----- 15 55 550 ** padraig23 is the current score leaderSo in this example, after taking everyone's score times 2/3 accuracy, and ranking them -- Elderhad would come out highest, then Charly , then the high accuracy --low score player. I would get a 50,000 point bonus for the idea and be in first place.Problem solved.I really don't see the advantage of the current formula where you have a percentile multiplied by a percentile.
Name------------ Score Accur. S*2/3(A)TMFBreakercharly 680 67.75 30713Elderhad-------- 860 66.09 37891padraig23**----- 60 100 4000Se7ensparks----- 15 55 550 ** padraig23 is the current score leader
Yet, the accuracy number does enable new players to experience some level of success right off the bat. This will likely be crucial to the long term appeal of the game / project. Maintaining that rating over time will be very difficult, especially if one regularly adds new picks.You raise a very good point here which, I think, goes to show just how difficult a problem this is going to be to solve. As a matter of fact, I can't personally think of a solution that doesn't leave at least something to be desired.Before I go on I want to make it clear that while I am identifying problems that I see in the player ranking system, on the whole, I think the current system is a pretty darn good one. Given all of the different and often conflicting goals including validity of results, simplicty and ease of use, ability of new players to have a fighting chance, etc., no system is going to be perfect on all of these fronts - and while I'm looking for ways to perhaps improve the current system, given all of the challenges, if I were grading the current system I'd give it an "A". My whole purpose here is to see if I can find a way to get us up into "A+++" territory.With the above on mind, one idea of an immediate 'fix' would be to increase the minumum pick list size. This would normalize the accuracy distributions a little bit more, and make it harder to get a good accuracy by sheer luck alone. While there may well be a better solution, this is in my opinion the best one I can think of at present. The problem with this approach, however, is that it doesn't solve the fundamental problem. The problem, if I'm right, is the 'linear vs. exponential' argument I made earlier in this thread. Increasing the minumum pick list size won't change that - so if we make the minimum pick list size 20 picks, all things being equal, the leader board should eventually be dominated by those who make close to 20 picks. In short, we won't have changed the fact that close to the minimum picks is the way to go, will only change the number of picks that this hovers around.Now, we could change the score vs. accuracy weightings. We could make score 3/4, accuracy 1/4, just as an example. This would help solve the more fundamental problem as it would delay the crossover point at which the exponential handicap to accuracy begins to outweigh the linear benefit to score - but it would also make it harder for new players to achieve a high rating. We wouldn't want new players to become quickly disillusioned as we'd lose their valuable contribution to community intelligence.We could also make accuracy a more 'true' measure - in other words do pretty much the same math I've done and measure the statistical significance of one's accuracy when taking into consideration one's pick list size. The potential problem with this approach is that lot of CAPS players who aren't particularly quantitatively oriented might have trouble grasping how this works - so this approach has a disadvantage on the simplicity, transparency, and ease of use fronts.This is a real dilemma. The way it stands now, CAPS has a bit of a problem when trying to foster the activity bias that helps create community intelligence. The problem, though, is that I can't think of a single 'fix' that doesn't create additional problems for CAPS elsewhere.Maybe some Fool a lot brighter than I am can come up with one - until then, however, I think increasing the minimum pick list size is the best option as it 'upsets the apple cart' the least. I'd also like to close by saying that I'm fully confident that the CAPS team will continue to think about this, and other ways to potentially improve CAPS, while keeping all of the challenging and often conflicting goals in mind. To the CAPS team specifically I'd like to say that I know this stuff isn't easy, especially while people like me keep pestering you with ideas/suggestions while you are trying to deal with a huge influx of new players, so to the CAPS team I'd just like to say, once again, that despite my occasional (hopefully constructive) criticism, my jester cap is off to all of you.Regards,Russella.k.a. TMFEldrehad
Russell,A thought.What about using volatility reducers in the player ratings. For example an overall rating would be a based on a 30 day moving average of one's score and accuracy.This will not alter the issue of dominance of smaller portfolios in the top 20, but will minimize those who only spend an hour or two there (like me last week). Plus players will progress more steadily up and down in rating, not drastically just because a holding went from green to red 10 minutes ago.For new players that have not built up the vested time to have a moving average, their page could display their current average and where it would rank if qualified (with disclaimer of time left for eligibility).I also like the earlier suggestion of initial portfolios being 7 stocks but require that it expand to at least 12 over a three month period. Not initially require too many to discourage new players from getting started, but eventually require enough to minimize small portfolio over-dominance. Combined with rolling average this should add some stability and consistency to the ratings, without discouraging new players.I haven't built out a Pascal's triangle for a 12 stock holding, but the statistics of having 83% (10 of 12) or better accuracy are likely quite low.ZzP.S. - Tech team, don't think I didn't see that eye roll.
Hi Charly. While I understand what you're saying about raising the minimum number of picks to 10 and requiring that 5 stocks be picked on an a rolling annual basis, I disagree. In my "real" portfolio, I own 5 stocks. I plan on holding them for awhile, especially EXEL, which I took full advantage of the fire sale and tripled my position. I think I've picked about 10 so far for this game. As the market changes, I will end some, and I will add some. I would think that there are a couple of solutions to this problem, and I'll explain why I feel that these solutions work well for all involved: My take is that this game is meant to mirror the player's actual investment strategy, or maybe try to get insight on successful strategies for a new investing strategy that they haven't tried before. As such, for myself personally in my actual stock portfolio, I own 5 stocks right now. I picked double the stocks that I would normally hold at any given time. Though it's not out of the realm of possibility that I would hold 10 if I had the necessary funds to meaningfully invest in 10 stocks, I don't so I won't. More than 10 is not gonna happen for me, and I love stocks as much or more than the next guy, but my personal strategy is to pick a small number of what I perceive to be winners by doing in-depth DD on each and every company that I consider as an investment, and investing only in those that make sense to me given all known information. #1) Forget the accuracy, it doesn't matter. A Rule Breaker would know this. You swing for the fences, and most of the time you miss, but when you hit it out of the park, all those misses are inconsequential. On the other hand, if you're right 90% of the time, but only marginally, that is also inconsequential. To avoid daytrading, people must have had a minimum of 7 picks for at least 30 days. Their initial ranking and returns are based off of a rolling 30 day % gain/loss. They must be in the game with at least 10 picks for at least a month before they can be considered for the overall top 10, but they can see where they would hypothetically rank every day were they eligible up until day 30. After day 30, they see where they actually rank based on their results. 2) My proposal is simple: minimum of 10 stocks, no maximum. Rankings are split into groups depending on how many stocks they have. 10-15, 16-20 etc. Rankings within each category are based solely on % gain. There can sure still be a top 10 list, based solely on % gain of their picks, but again, they must have 10 picks minimum for at least a month and their returns are calculated starting from their first 30 days results, though they are not eligible to be officially ranked for the first 30 days. This helps to alleviate "beginner's luck" and avoids have a different person atop the rankings every day. People just can't get that lucky if they're day-traders. I think enough info exists that points to day-traders under-performing the market, that it's not feasible to include them in this competition on the same terms and add their substantial commissions/trades in to the equation, as the spirit of the competition is investing for max returns. Day-traders should certainly have their own forum to compare results and should be ranked, but they shouldn't be allowed to be ranked in the overall top 10 for 30 days, again with their initial ranking ranking beginning with their % return over the 30 day waiting period. 3) The question that begs answering is how do you choose the best stockpicker out of all the different strategies that people have. I again think that rating people in seperate categories based on the number of stocks they choose is the clear answer. That gives all sorts of people from all different investing persuasions the opportunity to view the best stock pickers based on the number of stocks they choose to invest in at any given time. How do you choose a top 10 or 20 or 100 list overall? % return, plain and simple. They have to qualify by having at least 10 picks per day that run for a month straight. The picks can change daily, but they must have 10 total every day for 30 days straight. They can change their picks daily, or never change them depending on how they invest, but they must meet the standard minimum of at least 10 picks/day for 30 days, and their initial ranking when they become qualified for catagorical or overall ranking starts based on their last 30 days % gain/loss. I think that my proposition is a great solution for like-minded investors to connect and trade strategies, opinions based on the most basic choice in their investment decisions: how many stocks do you want to own? What do you look for in the stocks that are being frequently picked based on the number of stocks that you're choosing? How does the number of stocks you choose to invest in stack up against other numbers of stocks that others choose to invest in? Which strategy is working best at any given time? What are the characteristics of the common winners, regardless of how many stocks the the player holds, and what makes these stocks stand out above the rest? Answering these questions will help investors find the right strategy, and the winners of tomorrow. That is what this service is all about. Break it up into the number of stocks people hold, and offer the search functionality for them to do the rest. Break it up by sector, by size of the co, whatever, but give all participants a fair and common starting point.
I think David is right to say the system is too new to draw hard conclusions. For one thing I find it hard to believe that we'll continue to see new players with 7 random picks drift to the top much longer.First of all, at the moment the influx of new players is very large. But that will stabilize to a number equal or below to the number of new Fool subscribers. So the set of 'new' players will decrease sharply from now on. Among those new players, surely a certain percentage will have a high accuracy just by chance. However, if they also have a decent score, and especially when they can keep that score over some time, then you can wonder if it was luck.I think it's more likely that the top of the list will be dominated by 'old' players who made a small number of excellent picks that accumulated a good score over the S&P. No new player will be able to match that score in a short time by luck. 10-baggers take time to build, but if someone has one or two in a small list he'll be unpasable for new payers.Personally I think total score says more about a player's ability to pick good stocks than their rank. After all the scoer is already discounted with respect to the S&P500. So if the system gets adjusted such that players with a high score get ranked higher I'm all in favour. But I wouldn't throw out the dependence on accuracy alltogether, as I think that's a very important attribute of a good stock-picker as well. But I think the system must be made such that for a good stock-picker it's rewarding to keep adding picks. After all, one of the purposes of the system is to get a highly ranked player exert more influence over a stock's rating. That influence goes to waste if the top-ranked players just sit idly over their small number of picks and simply see them appreciate without ever picking a new one.Mark
Good lord...what kind of monster thread have I created?!? :)I completely agree with this statement:"P.P.S. - Before going 'live' it should be suggested that existing 'test' players be given the option of starting fresh."
I haven't built out a Pascal's triangle for a 12 stock holding, but the statistics of having 83% (10 of 12) or better accuracy are likely quite low.I'm not sure I remember Pascal's traingle, but I've been using the binomial distribution function in Excel to calculate the probabilities I've been using in this thread. I get the odds of going 10 for 12 or better as being about 2% (or 1 in 50). The odds of going 8/12 or better (a 75% accuracy) are between 1 in 13 and 1 in 14.Regards,Russella.k.a. TMFEldrehad
Pascal's Triangle represents the distribution of a 50:50 chance event propagating forward.http://mathforum.org/workshops/usi/pascal/images/pascal.hex2.gifFor example for 7 picks (row 8) the distribution of 7 for 7, 6 for 7, 5 for 7.....to 0 for 7 is:1 - 7 - 21 - 35 - 35 - 21 - 7 - 1. This sums to 128 possibilities.So the odds of being 7 for 7 are 1/128 = 0.78%. 6 for 7 or better = (1 + 7)/128 = 6.25%.The gif above goes to 10 picks.Honestly, Excel is easier.Zz
Sorry, a bit late on this discussion...Maybe it is due to my view from the bottom, but I think that you folks are trying to fix something that isn't broken. (hmm, my view from the bottom wouldn't make me say that)The top 20% of the players affect how the stocks are rated. Personally I think that accuracy should be counted as strongly as it is. Afterall, isn't the purpose of CAPS to find great stocks, not great players? The top 10 players list is a means to an end. That is our incentive to pick market beating stocks accurately. The player with the highest score may be the person whom I'd like to immitate in a real life porfolio if I were able to invest in say 25 different stocks every month. But if I am looking to invest in a single stock per month, then I want to take a closer look at the 5-star stocks and the folks who picked them.my $0.02KSW
KSW,Accuracy will be just as important as ever it was. I couldn't be happier that our original vision to ensure we're tracking both Accuracy and Score -- and to make a weighted composite be the Player Rating -- looks just as strong and relevant today as when we began developing this application and planning for this two years ago.The only "broke" thing that we're fixing, per Todd's excellent note and HelicalZz's excellent observation and suggestion, and strong community help from people like LoneIguana and TMFEldrehad, among others, is to make "Accuracy" itself more "accurate."With the proposed changes that we will be putting into play shortly, learned directly as a consequence of our private beta that you and many others have participated in and strengthened, we are making 15-out-of-15 more positively valued and weighted, more "accurate," than 7-out-of-7. The problem with our standings as it is, correctly foreseen some months ago by TMFEldrehad, is that they were disproportionately leaning to players who were making short pick lists because we were calling "Accuracy" just a percentage that didn't notice how many at-bats each hitter had taken, nor the probabilities of achieving or maintaining high Accuracy rates across large pick lists.I could go on for some length about how this one tweak helps us out in about four different ways, but I won't for now. Suffice it to say, to reply directly to the point you were making, that Accuracy plays no greater or smaller role as a consequence of this tweak. It plays the same interesting and critical role it has always played. All we're doing is making its very definition itself, thanks to the community intelligence, more... intelligent.Thanks for your ongoing participation in CAPS, for capping 33 stocks yourself so far, and for your enthusiasm and encouragement about A CAPSized World! :)Foolishly,David
Thanks for the very fast clarification David.I have to admit that I am at work and didn't take the time to thoroughly read each post. What I saw was a good detailed discussion, but one I thought wasn't taking the whole picture into account. I obviously was looking at the broader points and missed some crucial details.I was reading the various stat comparisons as a continuation of the score v. accuracy debate, not good accuracy v. great accuracy.So, well done. I'll go back to work now.KSW
It sounds to me that worries of a low number of picks leading to high rankings is greatly exaggerated. What is the goal of this game, to reward someone who picks 300 stocks? As it is now if two people have 50% accuracy and one picks 100 stocks and one picks 10 stocks, the one with 100 stocks will have a higher ranking. But who out there recommends picking 100 stocks? Good grief, if you pick 100 stocks and they each go up 1% more than the S&P you get 100 points, the same as if you pick 10 stocks that each go up 10 points greater than the S&P. Sure it is a lot harder to pick 100 stocks accurately, but even if you do, whose portfolio would you rather have, the one that has a lot of stocks and went up a total of 1% (boy can he pick 'em, we made 1%), or the one whose portfolio went up 10% but only made 10 picks? Isn't the goal to grow your portfolio? Although 100 points or 500 points or 1000 points is impressive, it is only impressive in light of the size of the portfolio with a smaller size reflecting a better stock picker.The second thing that has irritated me regarding this thread is all the emphasis on the 23% of people who make 7 random picks and have a score of 71. I think this was derived at by comparing people who make seven random picks against people who make greater than 7 random picks. The key here is random picks. So don't make random picks! Guess what, pick carefully and you will get a high rank. If all the new people make seven random picks, the number that make the 71st percentile will be substantially less than 23% if they are competing against skillful investors. I'm all for letting the people who got in early start over. But if you pick more stocks than you can realisticly manage, too bad. The game of investing is won by the one who plays the odds the best, not the one who plays the odds the most.
As it is now if two people have 50% accuracy and one picks 100 stocks and one picks 10 stocks, the one with 100 stocks will have a higher ranking.Actually, the way I understand the recently thawed mathematician's machinations, both of these players will hold the exact same percentile ranking in accuracy. The number of picks makes a difference at nominal accuracies above and below 50%, but at exactly 50%, how many picks one has doesn't matter one bit.I going to do my best to stay away from the geeky math, but the short answer is - it's harder to be 100% right over 100 picks than it is to be 100% right over 10 picks, and the new accuracy rules reflect that. With regard to being right 50% of the time, well, that's a 50/50 shot by blind guessing alone whether one has 10 or 100 picks - since both are equally easy, again, both will carry the same weight in accuracy under the new rules.Regards,Russell
With regard to being right 50% of the time, well, that's a 50/50 shot by blind guessing alone whether one has 10 or 100 picks - since both are equally easy, again, both will carry the same weight in accuracy under the new rules.That's exactly right, Russell. We assume an average accuracy of a little higher than 50 and it doesn't matter if you have 10 or 100 picks if you're right at the average accuracy. At the extremes, however, the number of picks is very influential on your accuracy rank. Todd
What is the goal of this game, to reward someone who picks 300 stocks? I have no input on the game design, but I think the answer to this question is: kind of. I think your complaints are based upon a premise that the game is designed to measure portfolio results. It isn't a portfolio game. The purpose of the game is to use community intelligence to rate stocks - thus, rules that encourgage investors to choose more stocks furthers the purpose. As a player - the goal is to be as accurate as possible and amass as many points as possible. Theoretically, the importance of accuracy and the 5% rule should limit the urge to pick as many stocks as one can type tickers for. Again, if you abandon the premise that this is a simulated portfolio game, you may change your mind as to the recent accuracy changes. T. Allan
Tar12 is right, I misunderstood the premise of the game. I thought the purpose was to identify players who pick successful portfolios in order for others to give increased weight to the top-pickers stocks. Furthermore, I haven't read the recent changes and will look for them.
We assume an average accuracy of a little higher than 50 and it doesn't matter if you have 10 or 100 picks if you're right at the average accuracy.May I ask what average accuracy you do assume? Or is that locked in the vault along with the plans to Regis?Regards,Russell
We used an average accuracy of around 52% because we feel that CAPS players will be slightly smarter than the market. The current global accuracy is hovering around 57% right now. My idea is that we can use the global accuracy of CAPS for our probability assumptions, but I haven't pulled the trigger on that one because we don't have a long enough time-frame yet. Todd