When can you be confident that you are a winner?

[jmillerdls]

How many picks? For the first time, i've been keeping track of all my pics this year (without betting actual money on them). I want to know at what point I can be comfortable investing my hard earned dollars on this and expect to come out on top in the long run. 200 picks...500, 1000?

Here are my rudimentary numbers from excel:



I do mostly money lines, but I always flat bet 1% of my bankroll (which I have kept a constant $110 unit). If I was doing this with real money, I would adjust my unit when my bankroll has grown 20%, and never adjust when my bankroll is down.

Because of my flat betting (the same $110 unit whether betting on the favorite, or underdog), and my most common bet being the moneyline underdog, I don't expect my winning percentage to be as high as others, I just simply expect to keep a good Return on Investment.

So, what is the generally accepted amount of picks to feel confident you are a winner?

Oh, and if anyone has any other stats I should be tracking in Excel, let me know...I just threw that all together.

[DeucesUp]

The answer to your question depends on a lot of factors. How certain do you want to be that you're a winner? What have been the terms of the bets you've made? How good have your results been? etc.


Without knowing more about the bets you made, I can't give you any exact numbers, but a few points:

First, your numbers and description don't quite add up. If you flat bet $110 on 374 bets and netted $2644, your ROI is no where near 14%, so you may want to check that, maybe I'm missing how you're calculating it.

For a point of reference, if all the bets you made were 50/50 propositions (bets with equal odds on either side) your win rate would be about 1.5 standard deviations above random, meaning there's about a 13% chance you'd achieve these results picking randomly. While the chance that you're picking better than 50% is quite good, you'd be much less certain that you're picking >52.4% that you'd need to beat standard vig.

Now if you're primarily picking ML dogs (as you apparently are), this changes things dramatically but you can't really say how much without know what the details of the bets made. Being up about 25 units after almost 400 bets is a starting point. As a side note, there was a thread about a month ago about what constitutes a "unit". I argued that generally one should bet "to win" 1-unit rather than risk 1-unit and this is a perfect example of why. By risking the same on all bets your results are heavily weighted towards how you did on big ML dogs and heavily under-weighted towards how you did on ML favorites. If you hit just a couple "extra" big ML dogs your overall results could be dramatically skewed to look better than they really are and conversely if you've been unlucky so far on big ML dogs and should of hit a couple more, your results might not be doing you justice.

Since these are imaginary bets anyway, I'd suggest you go back and change your bets all to win 1-unit, say $100 so a bet on a +300 dog would be $33 to win $100 and a -180 fav would be $180 to win $100. This will give a much more balanced view of how you're doing on all your bets.

As to whether you should start to bet real money, it's up to you. You've had very good results over a pretty large sample, it's certainly very likely you're a winner. I'd guess very few people establish this extensive a record before wagering real money. I think most books have limits as low as $5 some like the exchanges you can bet as little as you want. So if you're not comfortable putting much money at risk you can start out just about as small as you want and build up a bankroll. You can also churn through some bonuses which makes your risk even smaller.

[sparkywowo]

Going straight by the math we have the following:

ROI = Your Win % / Your Expected Win %

1.1414 = 0.54 / Your Expected Win %

So, Your Expected Win % = 0.47 (47%)

There is a Standard Error in Your Win % that depends on your sample size: err % = sqrt(Win% * Lose% / Sample Size)

err% = sqrt (0.54*0.46/374) = 0.026

So, your Win Percent with Error is 0.540 +- 0.025, while you need a Win Percent of 0.470 to break even.

0.540 - 0.470 = 0.070, a 0.070 difference,

0.070/0.025 = 2.8 Standard Deviations.

Looking this up on the Standard Normal Curve, we see that based on your current performance, the probability you at least break even on another representative sample is 0.997, 99.7%. So, I would state that mathematically, you have established your methods as viable, and if you continue to use them in the same way should continue to win.

50 picks is a small sample, 400 is decent. If you continue, try hard to stick with what you are doing and hang in there.

If you make 50 picks, with a 54% hit rate, you expect to hit 27 +- 4, between 23 and 31 of them, so there is about a 15% chance that your are even or behind after your first 50 money picks.

[BtBO]

Quick ROI check:

$41140 invested

$42420 returned

ROI = 1.03 Am I doing this right?

[jmillerdls]

You were right, the ROI was miscalculated (decimal in the wrong place). I personally don't adhere to the bet "to win" 1-unit philosophy. I feel this would greatly increase my variance and have me betting way above my bankroll when I take a favorite every now and then.
I still haven't decided what to do yet, but thanks for the advice. Good stuff.

[sparkywowo]

Given the lower ROI your advantage is marginally mathematically statistically significant, but too small to be practical and effective, and, you may even be overbetting your bankroll at 1% a bet.

[jmillerdls]

[ QUOTE ]
you may even be overbetting your bankroll at 1% a bet.

[/ QUOTE ]

This would be stunning to me. As everything I have read, and seen through my history has shown me it would be virtually impossible for me to go broke betting 1%. I'm sure it is possible, but it would require some major statistical imporobabilities to occur.

[DeucesUp]

[ QUOTE ]
I personally don't adhere to the bet "to win" 1-unit philosophy. I feel this would greatly increase my variance and have me betting way above my bankroll when I take a favorite every now and then.

[/ QUOTE ]

That's fine, but reweighting your picks would still be an interesting exercise because it will give a truer representation of how you're doing on all your picks. Note that you wouldn't be overbetting your bankroll to bet favorites to win 1-unit=1% but you ARE way overbetting your bankroll when you risk 1-unit=1% on big longshots, see below.

[ QUOTE ]
This would be stunning to me. As everything I have read, and seen through my history has shown me it would be virtually impossible for me to go broke betting 1%. I'm sure it is possible, but it would require some major statistical imporobabilities to occur.

[/ QUOTE ]

This again comes back to overweighting your dog picks. These calculations, if done properly, assume betting to win 1-unit. If you're just betting spreads it doesn't make any difference and as long as you're only betting ML's of +200 to +300 they're probably close enough, but if you bet +500 or +1000 or higher, your risk-of-ruin is very high.

A simple example:

Let's look at games where 1 team is a 10-1 favorite. Now you find 200 of these games where you like the favorite and 200 other games where you like the dog, let's see what happens:



Fav If the odds are set correctly, the fav will win 180 of these 200 games, this is what you need to break even. Even after the big vig charged, you still think you have a small edge and you expect to be able to pick "the right side" 183 times but it would not be unusual for variance to be 10 or more games over this 200 games sample, so betting 1% you're looking a loss of 7-units(7%) to a win of 13 units(13%). Not much profit, but very little risk.

Dog Now look at the other side, everything is the same. You need to pick 20 of 200 to break even, you expect to pick 23, and a variance of 10 games would not be unusual. Now you're looking at a range of +130units(130% gain) to -70units(70% loss).





Does this make sense? I hope this illustrates why when you risk the same on all bets, your overall results as well as your risk-of-ruin are dominated by how you do on the longshot bets. Now to be fair, I don't advocate a pure "to win 1-unit" scheme at extreme odds, you do have to scale it a bit. At 10-1 odds, I would wager 5 units to win 0.5 on a favorite and 0.2 units to win 2 on a dog.

Now you don't have to bet this way, a lot of serious bettors do just flat bet. I'm just saying it is difficult to access your success if you are flat betting over a wide range of odds, the longshots will always dominate.

[jmillerdls]

Fair enough, but I am much more willing to not be able to easily access my success than going broke. I think that kelly betting, and any system that varies your bet size is extremely volatile.

It is very rare for me to make a bet with +500 let alone +1000, so the chance of me going broke on these bets would probably take me the rest of my life. I think it would be much easier to go broke losing 20 bets where you have bet 5 units on a favorite than lose 100 units on betting underdogs...although clearly a ton of other stuff is involved in that.

I guess what I am trying to say is that if you are making some bets with 1 unit, and some bets with 5 units, then the only significant bets are the 5 unit ones...you might as well not even make the 1 unit bets.

Lets say you have a night where you have 4 favorites that you like. You will be risking 20% of your bankroll in one night. Am I the only one not comfortable with that? That seems like a vastly greater ROR than flat betting.

[DeucesUp]

[ QUOTE ]
Fair enough, but I am much more willing to not be able to easily access my success than going broke. I think that kelly betting, and any system that varies your bet size is extremely volatile.

[/ QUOTE ]

Wrong, flat betting over a wide range of odds is more volatile.


[ QUOTE ]
It is very rare for me to make a bet with +500 let alone +1000, so the chance of me going broke on these bets would probably take me the rest of my life. I think it would be much easier to go broke losing 20 bets where you have bet 5 units on a favorite than lose 100 units on betting underdogs...although clearly a ton of other stuff is involved in that.

[/ QUOTE ]

Wrong, it is easier to lose 100 units betting dogs, just look at the simpliest example of how likely it is to loose that many bets in a row:

Losing 20 bets in a row where you expect to win 90% = 0.1^20 =

.00000000000000000001

Losing 100 bets in a row where you expect to win 10% = 0.9^100=

.000027

Neither are likely, but losing 20 bets in a row as a 90% favorites is WAY more unlikely than 100 in a row as 10% dogs. You should be able to tell that this easily extrapolates to streaks where you the losses don't have to come in a row. It is hugely easier to go on a 100 bet downswing betting dogs than a 20 bet downswing betting favorties (all else being equal of course).

[ QUOTE ]
I guess what I am trying to say is that if you are making some bets with 1 unit, and some bets with 5 units, then the only significant bets are the 5 unit ones...you might as well not even make the 1 unit bets.

[/ QUOTE ]

Wrong, the significance is weighted by the payoff. A 0.2-unit bet at +500 to win 1-unit has the same significance as a 5-unit bet at -500 to win 1-unit. The example above should make that clear.


[ QUOTE ]
Lets say you have a night where you have 4 favorites that you like. You will be risking 20% of your bankroll in one night. Am I the only one not comfortable with that? That seems like a vastly greater ROR than flat betting.

[/ QUOTE ]

Yes, because my chance of losing 20% is 0.1^4 or 0.0001 or less.

If you find 20 +1000 dogs that you like, would you feel comfortable putting 20% of your bankroll at risk? Because your chances of losing 20% would be 0.9^20 = 12%