Evidence Bet

10+ years beating football bookmakers. Now I teach how to price a match yourself. No tips for sale — ever. 18+ · Gamble responsibly

[Evidence Bet @evidencebet]

1/3
I've been beating football bookmakers for over ten years. I stay anonymous on purpose.

I price fixtures across major and minor leagues every day to find mispriced odds. Here, I share the method so you can learn to price matches for yourself.

[Evidence Bet @evidencebet]

Live odds adjust in seconds after a goal (Croxson & Reade), and research links in-play betting with higher problem-gambling scores.

There is no edge in reacting live. Place your bets pre-match with your reasoning set, then just watch as a fan.

18+ · Gamble responsibly.

[Evidence Bet @evidencebet]

5/5 Takeaway:

• Size stakes so a 20-unit drawdown is boring
• Judge yourself by beating the closing line, not the last month
• Never raise stakes to win back a run

A solid process survives bad runs; impatience ruins bankrolls.

18+ · Gamble responsibly.

[Evidence Bet @evidencebet]

1/5 Even a genuinely profitable bettor spends far more time behind than you would expect.

With a true 51% vs 49% edge at even money (2.00), data from Joseph Buchdahl shows you have roughly a 25% chance of sitting in loss after 1,000 bets.

Variance is patient. 🧵

[Evidence Bet @evidencebet]

4/5 Bankroll recovery is brutally asymmetric:

• 10% drawdown needs +11% to recover
• 25% drawdown needs +33%
• 50% drawdown needs +100%
• 75% drawdown needs +300%

Staking is not about maximising profit on a good day. It is risk control to survive the mathematics.

[Evidence Bet @evidencebet]

Strikers don't carry "clinical finishing" across seasons. Research shows chance creation (xG/90) persists strongly year to year, but goals above xG barely carries over at all.

A hot finisher regresses to average. Price the volume of chances they get, not the streak.

[Evidence Bet @evidencebet]

Referees add extra stoppage time when a home side trails by one, and shorten it when they lead (Garicano et al.). Newer research shows club status plays a role too.

A trailing home side gets slightly more time to equalise. A small edge to keep in mind, not a bet on its own.

[Evidence Bet @evidencebet]

5/5
Takeaways for your pricing:

• Adjust draw probability upwards versus simple Poisson
• Know that low scorelines carry the bulk of that shift
• Never back draws simply because of a general slogan

Price the match properly first.

Save this for later.

[Evidence Bet @evidencebet]

1/5
If you use a simple Poisson model to price football matches, you are underpricing the draw.

Across roughly 40 Bundesliga seasons, draws happened about 10% more often than independent goal models predicted (Heuer & Rubner, 2012).

Goals are not completely independent 🧵

[Evidence Bet @evidencebet]

4/5
Does this mean bookmakers always underprice draws? No.

Across 479,440 matches (Kaunitz et al., 2017), consensus prices slightly overestimated draw rates, and only 2.1% of value bets landed on draws.

A single league might show a temporary edge, but it rarely lasts.

[Evidence Bet @evidencebet]

Risk of ruin with a +10% edge across 500 bets (Buchdahl simulation):

• All-in: 100%
• Martingale: 72%
• Fibonacci: 54%
• Flat stake: 13%
• 10% proportional: 0%

Even with an edge, chasing losses wipes you out. Stake sensibly.

18+ · Gamble responsibly.

[Evidence Bet @evidencebet]

5/5 The takeaway: take the lowest-margin price or the exchange line, strip out the margin, and treat that as your baseline. Use the soft bookmaker as where you look for value, never as your yardstick. Save this for later.

[Evidence Bet @evidencebet]

1/5 Before you can spot value, you need a reference price. But not every bookmaker's odds deserve to be your benchmark. To find out if a price is wrong, you first have to know where to find the price that is closest to right 🧵

[Evidence Bet @evidencebet]

4/5 A sensible caveat: the closing price of a sharp market is our best benchmark, not a flawless law. Market structures vary, and research shows occasional inefficiencies where opening prices or specific bookmaker dynamics retain small predictive traces.

[Evidence Bet @evidencebet]

Skill and luck look identical over short samples.

At odds of 2.00, the 95% noise band on ROI is ±8.8% after 500 bets and ±6.2% after 1,000.

Proving a +5% edge takes ~1,600 bets. A +5% ROI over 300 bets is just noise. Track CLV instead: it reveals skill far sooner.