[Data]

There is a very large number of bettors whom I will call here the followers. They may be following different things. It could be a winning handicapper or it could be a system that won ATS. They may realize they do not know if the capper will keep winning or if the lines will adjust to make the system unprofitable. So, they think that they will ride along while it wins and will stop as it starts losing. Sounds great but in practice it works just as opposite, they stick with the system while it loses and sit on the sidelines while it wins. Why? Here is why.

For simplicity sake let's call whatever they follow a system. Below is the illustration that shows what happens to the followers. Your numbers may vary. No followers were financially ruined while typing this post.

Let's say the system went 10-0 and the follower jumps on board. With the follower on board the system goes 2-8. The follower thinks the system is over and stops following it. The system goes 8-2 while the follower is not on board. The follower notices that the system is 20-10 and jumps on board again. The system goes 36-34. The follower loses on juice and, since the system have not made him any money in a long time and currently stands at unimpressive 56-44, he abandons it. The system goes 8-2.

Here are the results:
The system is 64-46 overall.
The system is 54-46 since the follower joined. Note, it actually won!
The follower is 38-42.

[roasthawg]

Regression to the mean is probably the number 1 reason why "followers" lose imo. As to your example you definitely have to be prepared for the ups and downs of sports gambling... especially the downs. No matter how good of a year I have I always have extended periods of time where it seems that I can't pick a winner... normally at least once or twice a year I see my bankroll cut in half in a relatively short time. Definitely gotta believe in what you're doing and hang in there.

[subs]

nice posts to read after 2 shockingly bad teaser weekends.

[Thremp]

What an utterly retarded post.

[TomG]

The followers lose because they make -EV wagers or wager too much on +EV wagers (mostly the former).

The concept of "regression to the mean" in a series of independent events is an illusion used to describe something that only exists in hindsight.

[Data]

The followers lose because they make -EV wagers or wager too much on +EV wagers (mostly the former).

The concept of "regression to the mean" in a series of independent events is an illusion used to describe something that only exists in hindsight.

I think you missed the point, must be the presenter's fault. This has nothing to do neither with the regression to the mean nor with EV. The point is that the followers manage to select -EV series of bets even if they have an +EV system to chose their bets from. This is rather a powerful observation that I do not claim to have made. I must declare that I borrowed (stole) it from the heavily quoted researches made by Dalbar, Morningstar, and Charles Ellis. Here is an example of a book quoting those: http://books.google.com/books?id=uXH...ngstar&f=false

[TomG]

Let's say the system went 10-0 and the follower jumps on board. With the follower on board the system goes 2-8.

Perhaps the system will go 8-2 with the follower on board. It's just as likely as going 2-8, assuming the system is just flipping coins.

Your whole post is one big Gambler's Fallacy.

[Data]

Perhaps the system will go 8-2 with the follower on board. It's just as likely as going 2-8, assuming the system is just flipping coins.

Your whole post is one big Gambler's Fallacy.

This is about TIMING. The system will go 8-2 and 2-8 number of times. The follower will jump in after "8-2" while missing the wins and will abandon after "2-8" after accumulating loses.

[Peregrine Stoop]

nice posts to read after 2 shockingly bad teaser weekends.

was it that bad? I thought 5 of 7 NFL legs covered their end?
last week was sad panda...
however, still been a plus year

[tomcowley]

I don't see how timing is relevant unless the EV of the bets follows a particular trajectory- more precisely if they keep following such a trajectory over and over, which takes quite a bit of imagination to find a way to apply to a single system or single capper on the timescale of tens of bets. If the EV of every bet is independent, long-term timing of noise will just change the results to EV/play*#plays*(%plays played) instead of EV/play*#plays.

The research, AFAICT from quick summaries, involves bouncing between funds that keep following those trajectories and buying in at a bad time. The alternative strategies don't seem analogous to options a non-originator has in sports betting.

[subs]

from my limited time spent following cappers - i have done exactly what Data is saying here (at least my understanding of it).

jumped on board when i think i have found a decent capper who then proceeds to have a bad run. only then to think this sucks and stop. if i had continued i probably would not have lost as much because stopping after a bad run makes my loses for that capper even worse. probably i have misunderstood? can't wait to get called a tard by thremp.

yea peregrine stoop i had only 4 NFLs with diamond and 2 of them lost + 1 low total college lost. so i hit 1 2 teamer . but the season did start well . i have 5dimes now and going to get wagerchief. if any1 is inclined to suggest another book then please do. sorry for digressing. just trying to really earn my tardship in full.

[Salamander]

This is a similar mindset to saying only bet roulette after 3 straight blacks come up, then bet red, since if you do this for long periods of time the reds must eventually catch up to the blacks. Like was posted earlier this is gambler's fallacy, the law of large numbers works for LARGE LARGE numbers.

[Pancho sanza]

I wouldn't totally discount what data is saying, its simliar to what people do in the stock market.

Peter Lynchs Magellen fund is an example, the fund returned 29 % over a 13 year period yet average investor returns were much lower because folks jumped in and out of the fund.

[Data]

I don't see how timing is relevant...

Care to make a sim? Let's simulate a "leader" (L) that is 54% and starts with 0-0 record, and a follower (F) who does not know the Ls rate. The F starts following once Ls record makes him 95% confident that the L is not an RNG but a winner. The F abandons the L if the L is on a losing stretch of any length that, if taken separately, makes an observer 95% confident that the L is not an RNG but a loser. Also, the F abandons the L if the L's overall record makes the F's confidence to fall to 68% or lower. The results after 100 bets should be very telling once the sim allows our F to follow 100 or 1000 Ls.

...unless the EV of the bets follows a particular trajectory- more precisely if they keep following such a trajectory over and over, which takes quite a bit of imagination to find a way to apply to a single system or single capper on the timescale of tens of bets. If the EV of every bet is independent, long-term timing of noise will just change the results to EV/play*#plays*(%plays played) instead of EV/play*#plays.

The EV may vary. The follower is guaranteed to miss winners which means he is missing a set of bets that likely have the largest +EV.

The research, AFAICT from quick summaries, involves bouncing between funds that keep following those trajectories and buying in at a bad time. The alternative strategies don't seem analogous to options a non-originator has in sports betting.

No, this research is about jumping in and out of a given fund. The "jumping" investors realize much smaller returns comparing to the fund's performance or its equivalent the "buy and hold" strategy.

[eonizuka]

I actually have seen this myself. I watched a guy give out baseball plays all season betting underdogs. In the end he finished with a losing record winning about 46% but because he was picking so many +150 and +200 dogs the system. Lots of times I would see people jump on after a few profitable days, lose a few then quit and miss the winning days.

[Thremp]

Wow. Just wow.

[luegofuego]

Errrrrr...

Seriously what the ****?

I don't think you realize what you're claiming in your OP Data because it's some amazingly misguided stuff. It's baffling that people that have an active interest in sports betting (was on my way to typing "winners", but edited myself for obvious reasons) doesn't understand the nature of streaks. You're actually saying that people that follow touts lose because they jump in at the end of hot streaks and jump off right when a hot streak is about to ensue, as if a pattern existed. Why not just follow every tout who has had a ten bet losing streak and stop when he inevitably hits ten good bets if there really is such an easily discernable pattern? Jesus ******* christ.

[tomcowley]

Care to make a sim? Let's simulate a "leader" (L) that is 54%

As long as the bets are independent 54%..

The follower is guaranteed to miss winners which means he is missing a set of bets that likely have the largest +EV.



No, this research is about jumping in and out of a given fund. The "jumping" investors realize much smaller returns comparing to the fund's performance or its equivalent the "buy and hold" strategy.

Sure- he's not maximizing ev because he's missing 54% plays or growth spurts (he's reducing his EV by the percentage of plays skipped, as I said). His strategy still isn't -ev compared to the benchmark of not betting at all unless you stipulate that the actual EV of the bets goes down predictably after a winning streak... and then comes back, and goes down, and comes back, and goes down, etc.

[Data]

Why not just follow every tout who has had a ten bet losing streak and stop when he inevitably hits ten good bets if there really is such an easily discernable pattern?

By doing so the follower will achieve a better winning percentage comparing to the tout's result. However, not neccessarily that is going to be above 50% and that is the reason not to do so.

[Data]

As long as the bets are independent 54%.

Sure, if you can do this that would be great. Even better if you have an idea how to make it 54% on average.

[Data]

Even better if you have an idea how to make it 54% on average.

Come to think of it, this could be a requirement as this is what happens in reality.

[tomcowley]

If you backtest a sample that IS 54%, then of course timing is actively reducing EV/bet because you're betting with more winners out of the deck and not betting when they're in there. This situation is analogous to a reverse counter in BJ. He's seen a lot of high cards, so now he bets. Because the final deck composition is constrained, he gets killed and does worse than random.

Obviously being "a 54% capper" in sports doesn't mean that you'll win exactly 54 out of the next 100 going forward, so the idea doesn't apply.

[aggieshawn]

I follow from way behind. Stop and turn around. therefore I lead from another direction.
Plato

[Data]

If you backtest a sample that IS 54%, then of course timing is actively reducing EV/bet because you're betting with more winners out of the deck and not betting when they're in there. This situation is analogous to a reverse counter in BJ. He's seen a lot of high cards, so now he bets. Because the final deck composition is constrained, he gets killed and does worse than random.

Was not this precisely my original point that the follower will realise a lesser winning percentage than the leader? On avearge, since the leader is unlikely to hit higher then 54% the follower is unlikely to hit higher than 50%.

[tomcowley]

Was not this precisely my original point that the follower will realise a lesser winning percentage than the leader? On avearge, since the leader is unlikely to hit higher then 54% the follower is unlikely to hit higher than 50%.

Going forward, the series of all bets and any subseries of those bets (chosen before the results of the bets being chosen are known) have to converge to the same long-term winrate as long as all the bets are independent.

[bztips]

Going forward, the series of all bets and any subseries of those bets (chosen before the results of the bets being chosen are known) have to converge to the same long-term winrate as long as all the bets are independent.

Exactly.

[Data]

Going forward, the series of all bets and any subseries of those bets (chosen before the results of the bets being chosen are known) have to converge to the same long-term winrate as long as all the bets are independent.

You assume that the edge is a constant. That assumptions is on one hand far-fetched and on the other hand is totally not needed. Since the edge varies, there is no single winrate for every bet but rather a distribution. A set of winning bets will have the bigger edges while the rest of the bets (the subset picked up by the follower) will have the smaller edges or no edge at all, or the negative edges.

[JohnnyC]

As always, good stuff Dad

[JohnnyC]

Oh this is "Data" not "Dad", in that case, gfy.

[tomcowley]

Unless there's a reason for bad bets and good bets to actively cancel out or to appear in equal numbers more often than pure chance would entail (like high cards and low cards in a shoe), it doesn't matter. Unless the EV actually changes somehow based on past results, it doesn't matter.

[luegofuego]

asdasfdsfds

[Data]

as long as all the bets are independent.

This seems to be the problem. Our sampling of games is without replacement.

[Lazyboy311]

This is about TIMING. The system will go 8-2 and 2-8 number of times. The follower will jump in after "8-2" while missing the wins and will abandon after "2-8" after accumulating loses.

Just my .02, but I think the above quote really sums up best what (I believe) the OP was meant to say. Disregarding individual chance, percentages, edges, and the like, the concept of people jumping on/off ship at certain times is very well documented when it comes to investing.

Take, for example, the most recent recession (call it a "losing streak"). Many, many people panicked, cashed out their 401k's, switched funds, hid their money in a jar in the yard...whatever. Their confidence level was so shattered that they felt no play in the market was better than where their money had been. However, the "correct" response was to continue buying at every price level while on the way down, lowering your overall average cost, and relying on the market to return to previous levels...when you would end up making MORE $$ because of your reduced cost (sticking with a system through thick and thin...or being the leader).

The alternative example would be the late-90's tech boom when everyone was buying into irrational p/e ratios, net loss companies, and the like (a solid recent record)...purely because things were on the up and up (a winning streak). But what ended up happening was that the market rationalized (a return to mean), prices fell, and people were left with overpriced junk stocks that weren't worth nearly what the irrational market said they were "worth" a year before. And then they sold, lost money, swore off investing, and missed the 03-06 boom (the next winning streak).

I know this is a very convoluted and possibly irrelevant example, but my only point is that a "normal" human being most often acts in ways that are counterproductive when it comes to risk adversion, whether it be in investing or wagering.

[tomcowley]

Uh.......... in backtesting, yeah, but that's not a meaningful concept when going forward since there's no defined finite future bet bin to be drawing from.

[Lazyboy311]

Uh.......... in backtesting, yeah, but that's not a meaningful concept when going forward since there's no defined finite future bet bin to be drawing from.

Right, but it's not about what the future IS as much as what an individual BELIEVES the future will be. People act largely based on the "what have you done for me lately" concept, which leads to making decisions based on the short-term past as opposed to a long-run view.