How we count expected wins on a results day
Expected wins is our own chances added up. How BAI8 uses it to judge a day of results, worked through with Tuesday's 13 real calls, and where it falls short.
On this page
Expected wins is our own chances added up. If we publish three calls at 50%, 60% and 70%, we expect 1.8 winners if those chances are right. On a results day we compare that number with the winners that actually came in. It answers one narrow question: did roughly as many of our calls win as our own numbers said they would? It does not tell you whether the day made money, and it does not prove anything over a single day.
The arithmetic, with Tuesday's calls
On Tuesday 20 October we settled 13 calls. Here are our published chances, straight from the record, grouped by whether the card had a positive edge.
| Group | Calls | Our chances | Expected wins | Actual wins |
|---|---|---|---|---|
| Positive edge | 4 | 47.2%, 51.0%, 60.2%, 44.7% | 2.03 | 1 |
| Negative edge | 9 | 48.1%, 52.8%, 53.8%, 48.9%, 53.3%, 48.3%, 68.6%, 42.6%, 67.7% | 4.84 | 6 |
| All | 13 | 6.87 | 7 |
Add the four positive-edge chances and you get 2.03. Add all thirteen and you get 6.87. Seven calls won. On this measure Tuesday was almost exactly on our numbers. The calls themselves are in Tuesday's 13 calls.
Why winning the right number can still lose money
Tuesday lost 2.03 units at the quoted prices, one unit per call, despite seven winners against 6.87 expected. Expected wins counts every winner as one, whatever its price. A Manchester City win at 1.08 pays 0.08 units; a Barcelona win at 2.60 would have paid 1.60. On Tuesday the short prices won and the longer ones did not.
That is why we show units next to expected wins in every results piece. The first number tests our chances. The second tests chances and prices together, which is what an edge is about. How the two relate is in expected value in betting in plain numbers.
What the gap means over a longer run
Across all 189 settled calls on the record, our chances add up to about 91.6 expected wins, and 117 calls have won. That gap is far larger than one day of luck. It does not mean the calls are better than we think; it suggests our chances may be too cautious overall, which would be a flaw, not a boast. It is a reason to check our calibration, which is what what a calibration curve shows is for.
Where it falls short
- It is our own yardstick. If our chances are wrong, expected wins is wrong in the same direction. It checks consistency, not truth.
- One day is noise. Thirteen calls can easily land two winners above or below expectation by chance alone.
- It says nothing about the price. A day can match expected wins and lose money, or miss it and make money.
What this does not tell you
Expected wins does not say whether the edges we publish are real. That is a question about prices, best answered by comparing our quotes with the closing line over many calls, as in why a closing line is the only honest scoreboard. It also cannot tell you what to expect from tomorrow's calls.
18+. Bet only what you can afford to lose.
Common questions
What does expected wins mean in betting?
It is the number of winners you should expect from a set of bets if your probabilities are right. Add up the chance of each bet winning and you have it.
How do BAI8 work out expected wins?
We add up our own published chance for every call settled that day. On Tuesday 20 October, 13 calls added up to 6.87 expected wins.
Why can a day beat expected wins and still lose money?
Because expected wins ignores price. If the winners are short favourites and the losers are longer prices, you can win more bets than expected and still lose units, which is what happened on Tuesday.
Does beating expected wins prove the model works?
No. Over one day the gap is mostly luck. It only says something when it holds over hundreds of calls, and even then it tests our chances, not whether the prices were beatable.