[SparJMU]

I have been refining a steam chasing method over the past few months in NFL, NCAAF, NBA, and NCAAB. My current record is 100-77. Based on the fundamentals of the system, and a 100-77 record, I am considering increasing my bet sizes significantly. I have two questions:

1) Is this sample size big enough to rely upon? How do I know if I am just getting lucky over a short period of time, or if I have really figured out a winning method?

2) I have been able to create subsets for my plays, one particular subset has hit at exactly 60%, the others hit around 54% and 56%. Can I use a Kelly calculator to determine my bet size on the various subsets, using 60% probability for my highest subset?

Obviously I am nervous to start risking big money on a steam chasing system, but if I can maintain these percentages I would love to make some decent money for my effort. Any advice is appreciated.

[SparJMU]

Anyone have experience with steam chasing systems? Is it reasonable to hit 60% at these rates or am I just getting lucky?

[Justin7]

That is a tiny sample size.

If you're placing a bet versus a stale number during a steam move, and beating the closer by a full point on sides, you're a favorite to win, but not because it's a system.

[Power Play]

1000+ plays is an adequate size. You can hit 55 percent over 100 by luck easily.

[JoeVig]

I would be interested if there are many places that would let you keep tagging their stale numbers at your significantly bigger bet size. Any A or B rated books involved?

[SparJMU]

Thanks guys. I kind of knew 177 games was too small, but was hoping that I was close. Getting to 1,000 games is going to take me a long time.

Definitely not on A+ books, but only one has reduced my limits so far.

[donjuan]

Your record is meaningless without attached prices unless you are only steam chasing spreads which obviously leaves a bunch of money on the table. ROI is more relevant.

[SparJMU]

My 100-77 record is entirely at -110. Not sure what you mean about leaving a bunch of money on the table.

[Tackleberry]

My 100-77 record is entirely at -110. Not sure what you mean about leaving a bunch of money on the table.

I believe he is reffering to the fact that if you are chasing steam and only betting the spread you are losing out on profitable opportunities by not also betting on the moneyline which should also be a soft number if the book is slow to adjust to what the market is saying regarding that game.

[SparJMU]

Interesting, thanks. Love the Tackleberry picture.

[Joey10]

I have been refining a steam chasing method over the past few months in NFL, NCAAF, NBA, and NCAAB. My current record is 100-77. Based on the fundamentals of the system, and a 100-77 record, I am considering increasing my bet sizes significantly. I have two questions:

1) Is this sample size big enough to rely upon? How do I know if I am just getting lucky over a short period of time, or if I have really figured out a winning method?

2) I have been able to create subsets for my plays, one particular subset has hit at exactly 60%, the others hit around 54% and 56%. Can I use a Kelly calculator to determine my bet size on the various subsets, using 60% probability for my highest subset?

Obviously I am nervous to start risking big money on a steam chasing system, but if I can maintain these percentages I would love to make some decent money for my effort. Any advice is appreciated.

Not sure how your system (Steam) works but my advice to you is start VERY low so you are not wagering a HUGE amount to win very little. Your sample size is good I guess my only question is that "Are they over multiple sports?" You really want to start with wagering on the same sport that you have data on.

Example of betting sizes - $10,000 bankroll should be a $10 -$25 wager assuming vigs are a -110 this way you do not feel like you are betting a lot on a game if you are chasing the 4th or 5th losing bet. Also How many wagers per day? I figure with your data you are looking at 4 -5 wagers a day. so this puts you WAY under 10% of your bankroll a day. The odds of losing ALL 4 -5 plays with a steam are rare but can happen and all the more reason not to go over board. The main reason people lose money is they do not know how to manage their bankroll. You figure if you were doing only $10 a wager (Unit) you would have made 100 X $10 = $1,000 (I am going by your record 100 -77) Since you are always looking to make the $10 on every play that is what you should have. Obviously the bigger the unit the more you would have made. The whole idea is that you can weather the bad streaks which will happen and continue to make money. The only time this hurts is when you raise your limits to fast.

With this method you really only need a 35- 40% win percentage.

I know you said it was just a steam play but with chasing you would have made $1,000 in my example and the regular betting you would have made $153 (based on a $10 wager and a vig of -110)

Hope that helped

[SportSharpener]

I would be interested if there are many places that would let you keep tagging their stale numbers at your significantly bigger bet size. Any A or B rated books involved?

Yeah, it seems that you can only make so much before the books catch on.

[That Foreign Guy]

If you're lucky the book looks at your $ win record and then makes a decision based solely on that. More intelligent books will see what numbers you are playing and kick you to the curb if / when they see you're always hitting stale numbers.



Joey, Steam chasing (waiting till a line moves and then betting at places that don't move quickly enough) is entirely different from stupid retardo chase progressive betting systems where you just bet more each time till you get in to profit. The first offers a chance of getting +EV bets in. The second just skews your results distribution to provide a lot of small wins and a few catastrophic losses.

[Peregrine Stoop]

I have been refining a steam chasing method over the past few months in NFL, NCAAF, NBA, and NCAAB. My current record is 100-77. Based on the fundamentals of the system, and a 100-77 record, I am considering increasing my bet sizes significantly. I have two questions: 1) Is this sample size big enough to rely upon? How do I know if I am just getting lucky over a short period of time, or if I have really figured out a winning method? 2) I have been able to create subsets for my plays, one particular subset has hit at exactly 60%, the others hit around 54% and 56%. Can I use a Kelly calculator to determine my bet size on the various subsets, using 60% probability for my highest subset? Obviously I am nervous to start risking big money on a steam chasing system, but if I can maintain these percentages I would love to make some decent money for my effort. Any advice is appreciated.

binomial test FTW dude
look at Wong's book for some simpler charts to extrapolate from

[kpoutlaw]

Hey, I paid Sportsinsights $200 a few months ago to follow steam moves myself but didn't really follow through and cancelled after only a month. Do you use it too to follow steam moves?

[specialronnie29]

binomial test FTW dude
look at Wong's book for some simpler charts to extrapolate from

no one has answered the guys question yet

it is on the border of being statistically significant with a 95% confidence level

http://stattrek.com/Tables/Binomial.aspx

use 177 trials, 100 successes, true prob 0.5

[SparJMU]

Peregrine and Ronnie, thanks for the tips. It looks like I am closer to a significant sample than I thought.

[specialronnie29]

so moral of the story is dont listen to justin7

[Pokerjoe]

no one has answered the guys question yet

it is on the border of being statistically significant with a 95% confidence level

http://stattrek.com/Tables/Binomial.aspx

use 177 trials, 100 successes, true prob 0.5

You're calculation is correct. You're math is terrible. Better to listen to J7 after.

[JustinBieber]

According to you steam chasing isn't +EV anyway.

[SparJMU]

All right I am a little confused now. So I am 100-77 and have made some good money, but I am considering increasing my wagers significantly. After looking at that website from Ronnie I thought my sample size might be significant, but then again I have zero experience with binomial distribution.

Perhaps the original question was a dumb one. I am just trying to figure out if my experiment is something I can rely upon and expect continued success.

[bztips]

Not sure what else you want to know. As Ronnie showed, you're right at about 5% significance level, meaning there's a 5% chance the results you've gotten are just by luck (ie., hitting 100 out of 177 even though your true probability of winning each bet is only 50%). It's up to you to assess whether that's good enough to start making larger wagers.

[SparJMU]

OK Thanks. I think its time to let this thread die. Thanks for all contributions.

[Pokerjoe]

According to you steam chasing isn't +EV anyway.

If this is addressed to me, you're wrong. I've said that steam chasing isn't for me, because I don't personally want to sit in front of a computer all day, but not that it isn't +EV for anyone. It is +EV when done well.

My stance about the misunderstanding (not the error) of BTCL is not evidence to the contrary.

[Pokerjoe]

Not sure what else you want to know. As Ronnie showed, you're right at about 5% significance level, meaning there's a 5% chance the results you've gotten are just by luck (ie., hitting 100 out of 177 even though your true probability of winning each bet is only 50%). It's up to you to assess whether that's good enough to start making larger wagers.

Bztips, I think your summation, no disrespect, is the problem. Math applied without philosophy is mere calculation, and nearly worthless (although the reverse is not true).

Consider a system/approach, whatever, as random.
Maybe the last 19 guys who tried it during different stretches failed, said "meh," and didn't post about it in a forum.
1 guy at random hits a stretch of results where it "works", which it should at random sometimes do. He does post here and is told that there is only a 5% chance he's been lucky.

[JustinBieber]

If this is addressed to me, you're wrong. I've said that steam chasing isn't for me, because I don't personally want to sit in front of a computer all day, but not that it isn't +EV for anyone. It is +EV when done well.

My stance about the misunderstanding (not the error) of BTCL is not evidence to the contrary.

You claimed BTCL is not +EV.

BTCL = Steam Chasing.

[bztips]

Bztips, I think your summation, no disrespect, is the problem. Math applied without philosophy is mere calculation, and nearly worthless (although the reverse is not true).

Consider a system/approach, whatever, as random.
Maybe the last 19 guys who tried it during different stretches failed, said "meh," and didn't post about it in a forum.
1 guy at random hits a stretch of results where it "works", which it should at random sometimes do. He does post here and is told that there is only a 5% chance he's been lucky.

So by your logic no one who finds a result that's "statistically significant" should bother posting, because obviously they're the 1 out of 20 that's just found random good luck. Right.

[Data]

I have been refining a steam chasing method over the past few months in NFL, NCAAF, NBA, and NCAAB. My current record is 100-77. Based on the fundamentals of the system, and a 100-77 record, I am considering increasing my bet sizes significantly. I have two questions:

1) Is this sample size big enough to rely upon? How do I know if I am just getting lucky over a short period of time, or if I have really figured out a winning method?

2) I have been able to create subsets for my plays, one particular subset has hit at exactly 60%, the others hit around 54% and 56%. Can I use a Kelly calculator to determine my bet size on the various subsets, using 60% probability for my highest subset?

Obviously I am nervous to start risking big money on a steam chasing system, but if I can maintain these percentages I would love to make some decent money for my effort. Any advice is appreciated.

It all comes down to whether you bet a stale number that has +EV based on the market or you bet a widely available number after the line move has occurred. In both cases your record is totally irrelevant. In the first case you should size your bets based on the EV of the given number. In the second case your EV is most likely 0 and the bet size is 0 as well, no matter how big is your sample size. Since the "process" (system) is based on evolving factors which nature is unknown (the lines) there is no way to know when and how the "process" and its winning rate change. All those significance tests assume that the process remains the same and therefore are not applicable here.

Furthermore, even if the process currently has a winning rate and by itself is not 0EV but +EV it is still most likely of no help due to, if you allow me, my trademark "followers dilemma":

...my stated belief that it can only lead to losses. That belief is based on the simple idea I presented here: http://www.sportsbookreview.com/forum/handicappe...kely-lose.html.

Your strategy is just another example of a huge array of strategies where the bettor follows a strategy (a system, a tout, a black box) that was successful in the past without knowing if its edge is still there. The problem the follower will face over and over again is what should he do after a series of losses (stopping vs. continuing). Case in point: http://www.sportsbookreview.com/forum/players-ta...-bad-runs.html. The only meaningful answer to the follower's dilemma is to stop following the system after its first loss. Obviously, this affair will be very short. Otherwise, sticking with the system is likely a better approach than jumping in and out but it too guarantees losses due to staying with the system that no longer has the edge for too long.

[bztips]

It all comes down to whether you bet a stale number that has +EV based on the market or you bet a widely available number after the line move has occurred. In both cases your record is totally irrelevant. In the first case you should size your bets based on the EV of the given number. In the second case your EV is most likely 0 and the bet size is 0 as well, no matter how big is your sample size. Since the "process" (system) is based on evolving factors which nature is unknown (the lines) there is no way to know when and how the "process" and its winning rate change. All those significance tests assume that the process remains the same and therefore are not applicable here.

Furthermore, even if the process currently has a winning rate and by itself is not 0EV but +EV it is still most likely of no help due to, if you allow me, my trademark "followers dilemma":

Can't agree with you here Data.

IF you believe, as some do, that by definition the closing line is always the most efficient line, then the process is NOT based on evolving factors that should affect your bets. In other words, no matter the underlying evolving factors, if the closing line is still the efficient line and if you can beat it then you should bet. The OP's historical record is in fact a test of this hypothesis.

And just for the record, even though his results are on the border of statistical significance, personally I'm not an "all-in" believer in the BTCL theory, because I don't believe that it's necessarily fully efficient for every specific event, even though it probably is "on average", at least in large liquid markets. (And I know you're even less of believer than I am.)

[Data]

Can't agree with you here Data.

Perhaps, this is due to my poor presentation. What you said below is not what I was talking about.

IF you believe, as some do, that by definition the closing line is always the most efficient line, then the process is NOT based on evolving factors that should affect your bets. In other words, no matter the underlying evolving factors, if the closing line is still the efficient line and if you can beat it then you should bet. The OP's historical record is in fact a test of this hypothesis.

The evolving factor is not the intra-day line moving but (1) the process of linemaking and (2) the market participants that create the steam. A change of any of these two factors deem the record irrelevant and that change is undoubtly taking place.

[bztips]

OK, gotcha. That I agree with.

[specialronnie29]

Bztips, I think your summation, no disrespect, is the problem. Math applied without philosophy is mere calculation, and nearly worthless (although the reverse is not true).

Consider a system/approach, whatever, as random.
Maybe the last 19 guys who tried it during different stretches failed, said "meh," and didn't post about it in a forum.
1 guy at random hits a stretch of results where it "works", which it should at random sometimes do. He does post here and is told that there is only a 5% chance he's been lucky.

this is funny because you have just fallen into exactly the trap you are discussing

the guy wanted to know what was HIS expectation of being 'onto something'. And I answered his question.

your question is what should be OUR expectation of him being 'onto something', which is different for the reason you have mentioned.

the difference is between Prob(Being onto something) [the first case] and Prob(Being onto something|those who don't get winning records don't post about them) [the second case]