[voros]

The science of probability was first developed by problem gamblers looking for an edge.

In tribute to one of Mathematics' finest hours, I figured I'd use the gambling world to try out my new international ratings and see how it does.

Today the African Cup of Nations begins in Tunisia. Important for my purposes because most of the online bookmakers have odds posted for the matches.

The idea is to compare the ratings to those of a set of experts (in this case the bookmakers and moreso the degenerates that engage in this activity that is illegal in most of the US. )

I'll place a series of mythical bets from a percentage of a mythical stake, on the odds posted. I'll also keep track of the bets with a second set of odds, which are the same odds, but with the house's "juice" removed. My general theory on sports gambling is that you can beat your fellow gamblers, but not the "juice."

Anyway, to the meat:

Tunisia (system ranks 49th in the world) vs. Rwanda (ranked 122nd) in the opening Match in Tunisia. Odds:

Result         Odds     Fair Odds
Tunisia Wins    1/4          2/5
Draw            7/2          4/1
Rwanda Wins     9/1         10/1

The following is important:

PLEASE, NO ONE WAGER ANY ACTUAL MONEY BASED ON THE FOLLOWING. Thanks.

The system says that Tunisia should win roughly 84% of the time. Some easy math shows that this rises above the even bet threshhold for each set of odds, so:

Stake = $2,000
Bet (5%) = $100

It's $100 on Tunisia to win today's match versus Rwanda. That'd be a profit of $25 bucks if they win or with the "fair" odds $40.

More matches tomorrow. Hopefully I keep up (though the spreadsheet is set up, so this typing part takes longer than anything).

[NoSix]

Originally posted by voros
My general theory on sports gambling is that you can beat your fellow gamblers, but not the "juice."

I hold similar views - but am interested to see how you do. Tunisia 2, Rwanda 1 - so far so good!

[voros]

Tunisia was a heavy favorite, system was just surprised they weren't heavier favorites:

Three games Sunday:

Bet - Odds (Fair Odds)
Guinea (67th) win over Dem Rep of Congo (81st) - 2.6 (2.93)
Egypt (47th) win over Zimbabwe (79th) - No Bet (1.94)
Algeria (66th) win over Cameroon (26th) - 6.5 (7.3)

System says Egypt v. Zimbabwe is pretty well handicapped, and there is no even bet (but making the odds "fair" makes Egypt a bet).

I'm going to halve the bet for Algeria and other underdogs like it, so as to be able to manage risk a bit, but still play them. System thinks Cameroon should win, but gives Algeria a better chance than the odds do.

[voros]

This thread needs an "[R] - International Results" in the title, but I can't edit it. Any help?

[voros]

Okay, African Cup of Nations Results below:
.
.
.
.
.
.
.
.
.
.
.
.
.
Guinea 2 - 1 Dem Rep of Congo (Win on both)
Egypt 2 - 1 Zimbabwe (lay off, win on "fair" bet)
Algeria 1 - 1 Cameroon (lose longshot on both)

Algeria certainly gave it the college try for us. The win with Guinea at better than even money helps.

The bets were $101 on the first two and $51 on the last.
Current stakes: Stake: +$135.60; "Fair" Stake: +$281.74

Monday's bets:
Kenya (91st) win over Mali (68th) - 5.00 (5.63)
Senegal (30th) win over Burkina Faso (80th) - No Bet (1.73)

The Kenya longshot bet is halved.

The system says that the Senegal match is handicapped fairly well, so there's no bet on the real odds, but a Senegal bet on the fair odds.

[beineke]

Interesting study ... perhaps you'd be willing to add a bit more data. Would you mind including your model's aggregated "expected winnings?"

It'd be interesting to know if your model thinks you've been having good luck or bad luck.

[voros]

Originally posted by beineke
Interesting study ... perhaps you'd be willing to add a bit more data. Would you mind including your model's aggregated "expected winnings?"

It'd be interesting to know if your model thinks you've been having good luck or bad luck.

At this early stage, I'm clearly having good luck. Most of the bets only have an expected return of under 10% per bet.

Though the good luck may be ending as I lost both of Monday's bets:

Mali 3, Kenya 1 (loss for both)
Senegal 0, Burkina Faso 0 (loss for fair odds, no bet on real odds)

Interesting bet today as the system says to go with a draw:

Morocco draw with Nigeria - 3.00 (3.38) (halved bet)
South Africa win over Benin - 1.40 (1.58)

If neither comes in, I will be behind in both sets of odds.

I'm halving any bet that the system says isn't the favorite to come in (system says a Nigeria win over Morocco should be a slight favorite over a draw).

Stake = $2,082.60, Fair Stake = $2,110.74

[beineke]

Originally posted by voros
At this early stage, I'm clearly having good luck. Most of the bets only have an expected return of under 10% per bet.

Very rough estimates ...
E(winning per bet) = $50
Std. dev. (winning per bet) = $1000

After 32 games, expected winnings are $1600, but with a standard deviation of nearly $6000. Based on these numbers, if the rating system is correct, there is about a 40% chance of ending up behind. And even if the expected winnings were doubled to $100 per bet, there would be a 30% chance of ending up behind.

To test a system like this, it looks like we need an awful lot of games.

[NoSix]

Originally posted by beineke
Very rough estimates ...
E(winning per bet) = $50
Std. dev. (winning per bet) = $1000

After 32 games, expected winnings are $1600, but with a standard deviation of nearly $6000. Based on these numbers, if the rating system is correct, there is about a 40% chance of ending up behind. And even if the expected winnings were doubled to $100 per bet, there would be a 30% chance of ending up behind.

To test a system like this, it looks like we need an awful lot of games.

It seems to me if you are to have any chance of beating the odds, you would need to adopt a system where the amount of the bet increases as the difference between your predicted odds and the offered odds increases (e.g., Kelly's formula).

[microbrew]

Who said gambling can't be honest hard work? In theory...