Learn
The Long Run Test: positive EV betting variance in one table
Positive EV betting variance is why good bets still lose: in a hypothetical bet that wins half the time at +105, worth about +2.50 dollars per 100, a hundred bets leave you down about 40 percent of the time, a thousand about 22 percent and five thousand about 4 percent. The tables below show those odds for three edges, and how much the edge shrinks if your fair chance is off by a point or two.
The long run test is a one-page table that shows how many bets it takes before a real edge shows up in your results. It is the table from our video on positive EV betting, with the numbers worked out so you don't have to.
The video's example, step by step:
- A hundred bets. A hypothetical bet that wins half the time at +105 is worth about +2.50 dollars per 100 on average, and a hundred of them still leave you down about 40 percent of the time.
- A thousand bets. The chance of being down falls to about 22 percent, roughly one time in five.
- Five thousand bets. Only now does a loss get rare, at about 4 percent.
- The fragile edge. If the real chance is 48 percent instead of 50, the same bet is worth about minus 1.60 dollars per 100.
Below are the full tables, the math behind them and how to use them on your own bets.
Why good bets still lose
Expected value is the average result of a bet if you could repeat it many times: each outcome times its chance, added up. It is a long-run number, and the law of large numbers is what makes the average show up, but only over many trials. Over a short run, the spread of results is much bigger than the edge.
Take the video's hypothetical. A bet wins 50 percent of the time and pays +105. Using the expected value formula, a 100 dollar bet is worth 0.5 times 105 minus 0.5 times 100, which is +2.50 dollars. But each single bet either wins 105 or loses 100, so the swing on one bet is about 102.50 dollars, around forty times the edge. That gap is why a real edge can hide for months.
The table: chance you're down after N bets
Every number below is hypothetical and assumes flat 100 dollar bets on outcomes with a true 50 percent chance, at three prices. The chance of being down is worked out with a normal approximation from the expected value and its spread, so treat it as close, not exact.
| Bets placed | At +102 (EV +1.00 per 100) | At +105 (EV +2.50 per 100) | At +110 (EV +5.00 per 100) |
|---|---|---|---|
| 100 | 46.1% | 40.4% | 31.7% |
| 200 | 44.4% | 36.5% | 25.0% |
| 500 | 41.2% | 29.3% | 14.3% |
| 1,000 | 37.7% | 22.0% | 6.6% |
| 2,000 | 32.9% | 13.8% | 1.7% |
| 5,000 | 24.2% | 4.2% | under 0.1% |
How to read it: find the column closest to your typical edge and the row closest to the number of bets you've placed. With a 2.5 percent edge and 200 bets behind you, there is still more than a one-in-three chance you'd be down even if every bet was a good one. With a 1 percent edge, five thousand bets still leave about one chance in four.
The fragile edge: what if your fair chance is off
The table assumes the 50 percent is right. In practice it is your estimate, and small errors matter more than they look. At +105, here is what one 100 dollar bet is worth at different true chances, using the same expected value math (hypothetical):
| True chance | EV per 100 at +105 |
|---|---|
| 52% | +6.60 |
| 51% | +4.55 |
| 50% | +2.50 |
| 49% | +0.45 |
| 48% | -1.60 |
| 47% | -3.65 |
The break-even chance at +105 is 100 divided by 205, about 48.78 percent. So the whole edge in the video's example rests on the fair chance being more than about 1.2 points above break-even. That is why the fair price you start from matters as much as the price you get.
How to find the fair chance in the first place
The simplest starting point is the market itself. A two-way price includes the book's margin, which the mathematics of bookmaking calls the overround: the amount by which the implied chances add up to more than 100 percent. To remove it, convert both sides with the standard moneyline conversion and divide each by the total.
For example, a hypothetical market at -120 and +100 implies 54.55 and 50 percent, a total of 104.55 percent, using the moneyline conversion. Dividing by the total gives fair chances of about 52.17 and 47.83 percent. If another book offers the second side at +115, that bet is worth about +2.83 dollars per 100 at those fair chances.
Starting from the market is a reasonable default because the market is hard to beat. A PLOS ONE study found that ratings built from betting odds outperformed the other rating systems it compared for soccer. Some researchers argue a model can still profit if its errors are different from the market's: Hubacek and Sir describe beating the market with a less accurate model that is decorrelated from market prices. Either way, a fair chance needs a reason, not a hunch.
How much to bet while the long run plays out
Because the swings are large, bet size decides whether you survive long enough for the edge to show. The Kelly criterion gives the growth-maximizing fraction of a bankroll: f equals p minus q divided by b, where b is the net odds. For the video's example (p = 0.5, +105, so b = 1.05) that is about 2.38 percent of the bankroll, and half Kelly is about 1.19 percent.
Two warnings from the same source. Betting more than Kelly raises the risk of ruin, and overestimating the win chance pushes the Kelly bet above the right size. Since your fair chance is an estimate, many bettors use a fraction of Kelly, which reduces volatility and ruin risk at the cost of slower growth.
How to use the long run test on your own record
- Count your bets and estimate your average edge from the prices you took against fair prices, not from your results.
- Find your row and column in the first table. That is roughly how often a bettor with your edge would be down at this point by luck alone.
- Judge the process, not the scoreboard. If you are inside the normal range, a losing record says little. Keep checking that your prices beat fair prices.
- Recheck your fair chances. Use the second table: if your fair estimates are a point or two too optimistic, the edge may not be there at all.
The positive EV betting guide covers finding +EV prices step by step, and the expected value calculator runs the math for any price and chance.
Quant+ is analytics for informational purposes. Nothing is guaranteed. 21+. If you or someone you know has a gambling problem, call 1-800-GAMBLER.
Questions people ask
Can you lose money betting positive EV?
Yes, often over short runs. In the hypothetical at +105 with a 2.5 percent edge, about 40 percent of bettors are still down after 100 bets, and about 22 percent after 1,000.
How many bets does it take for positive EV to show up?
It depends on the edge. With a 2.5 percent edge, the chance of being down falls to about 4 percent after 5,000 bets; with a 1 percent edge it is still about 24 percent at 5,000.
What happens if my fair probability is wrong?
The edge can vanish. At +105, a true chance of 50 percent is worth +2.50 per 100, but 48 percent is worth about -1.60 per 100, because break-even at that price is about 48.78 percent.
How much should I bet on a positive EV bet?
A small fraction of your bankroll. Full Kelly for the +105 example is about 2.38 percent and half Kelly about 1.19 percent; many bettors use a fraction because their fair chances are estimates.
Sources
- Expected value, Wikipedia
- Mathematics of bookmaking, Wikipedia
- Moneyline odds, Wikipedia
- Kelly criterion, Wikipedia
- The Betting Odds Rating System, PLOS ONE (Wunderlich, Memmert)
- Beating the market with a bad predictive model, arXiv (Hubacek, Sir)
