The Staking Secret Sharp Bettors Don't Talk About

The luckiest I ever got wasn't a winner at long odds. It was stumbling across an academic paper.

I was deep in a professional betting run: researching edges, building confidence in my selections, finding spots where I genuinely believed I had the bookmaker beat. My results were good. Not great. Good.

Then I found a paper by two economists at the Universidad Complutense de Madrid (Barge-Gil & García-Hiernaux, 2019) on staking methods for bettors who know they have an edge but can't precisely calculate how much of one.

That was me exactly.

I took my own betting history and ran it through their proposed method, something they called "unit impact", as a backtest against how I'd actually been staking. The difference was significant enough that I changed my approach that same day.

My ROI climbed. Compounding kicked in harder. And looking back, it was the single biggest upgrade to my betting results that didn't involve finding a better bet.

The rest, as they say, is history.

Here's everything you need to understand about it, and why it's backed by 55,891 bets of hard evidence.

Kelly: The Gold Standard (That Most Bettors Can't Use)

In 1956, Bell Labs scientist John Kelly published a paper that changed gambling and investing forever.

His formula, now called the Kelly Criterion, tells you the exact percentage of your bankroll to bet on any given wager to maximise long-term growth (Kelly, 1956). Not this week's growth. Not this season's. Long-term, compounding, mathematically optimal growth.

Ed Thorp used it to beat blackjack. Then to beat the stock market. Warren Buffett has spoken about its principles. It's the only staking framework that's never been mathematically bettered (Maclean, Thorp & Ziemba, 2010).

The formula is straightforward:

Stake % = (p × odds − 1) ÷ (odds − 1)

Where p is your true probability estimate and odds is the decimal price.

The catch? You need an accurate p.

And I mean genuinely accurate. The kind of accuracy you get from deep historical modelling, thousands of data points, and serious quantitative work. Professional quant funds use it. Algorithm-driven betting syndicates use it. The rare human who can assign calibrated probabilities with real precision can use it.

Most punters cannot.

Not because they're bad at betting. Because honest probability estimation is extraordinarily hard. If you think Carlton kicks 90+ against Collingwood on Saturday and you put that at 62%… that's a feeling, not a probability. Kelly doesn't work on feelings. Fed a bad probability, it either over-bets and blows your bankroll or under-bets and leaves money on the table.

So what do you do?

The Three Staking Plans Every Bettor Defaults To

When Kelly is too demanding, bettors fall into three camps.

1. Unit Loss (Flat Staking)

Risk the same dollar amount every bet. $50 on every tip, regardless of odds.

It's simple. It's the default. And around 60% of professional tipsters on major platforms use it (Barge-Gil & García-Hiernaux, 2019).

The problem: a $50 bet at $1.50 odds and a $50 bet at $8.00 odds are not the same bet. If the long-shot hits, it swings your bankroll dramatically. If it misses, it barely moves. Your bankroll becomes a hostage to your biggest-priced bets.

2. Unit Win

Bet whatever amount wins you the same profit every time. Want to profit $50? At $1.50 you risk $100. At $8.00 you risk $7.14.

Short-priced bets now absolutely dominate your risk. One bad run at short odds and you're gone.

3. Unit Impact

Bet so that the difference between winning and losing is identical for every bet, regardless of price.

This is the one most punters have never heard of, and it's the one the research supports.

What the Research Actually Shows

Barge-Gil & García-Hiernaux (2019) tested all three methods using 55,891 bets from professional tipsters on Pyckio, one of the world's largest tipster platforms.

Their core question: which staking method best approximates what Kelly would recommend if you did know your true probabilities?

The results were decisive.

Unit Impact matched the Kelly framework far better than either flat staking or unit win. Statistically, the relationship between expected yield and price that Unit Impact assumes (yield scales with (odds-1)/odds) matched real professional betting data much more closely than the linear relationship flat staking assumes.

Then they ran the bankroll simulation.

Starting from the same point, with the same bets, using optimally calibrated stakes:

• Unit Impact produced the best final bankroll
• Unit Loss (flat) produced a bankroll 5x smaller
• Unit Win was so catastrophic it was essentially meaningless. The Unit Impact bankroll was 16 million times larger

Sixteen million times.

That's not a rounding error. That's the difference between a staking method that respects the structure of odds and two methods that ignore it.

My Method: 5% ÷ Odds

After reading the paper and backtesting my own history, the practical formula I landed on was simple:

Stake = 5% of bankroll ÷ decimal odds

At $2.00, you're betting 2.5% of your bankroll. At $4.00, you're betting 1.25%. At $1.50, you're betting 3.3%.

It isn't pure Kelly. I couldn't estimate true probabilities with that precision and neither can most people. But it's a clean, practical approximation of Unit Impact that scales your exposure sensibly with price without requiring a PhD in statistics.

I also used this when evaluating tipsters in sports I wouldn't research myself. I'd take their full track record and run two simulations side by side: their stated staking versus my simple 5 ÷ odds approach.

Every single time, my method outperformed theirs, except for the rare few who were already doing something similar.

The math wasn't magic. It was structure. Letting each bet have proportional impact on the bankroll rather than letting long-priced bets dominate through flat staking or short-priced bets bleed you out through unit win. That single change, one formula applied consistently, is what made the compounding work properly.

One Critical Warning: Don't Look Like a Robot

There's something nobody in staking theory talks about: bookmaker account health.

If you're staking $12.47 on one bet and $8.93 on the next, you look like a machine. Bookmakers, particularly Australian retail books, are extremely alert to systematic, calculated stake sizing. Accounts that bet with unusual precision get flagged, limited, or closed.

The mathematics might say $12.47. But you bet $12 or $13.

Give up a tiny fraction of theoretical precision to look like a normal human being. Round to the nearest whole dollar. Sometimes round to the nearest $5. Vary slightly. The edge you lose is negligible. The account longevity you gain is worth it.

Sharp bettors know the game isn't just finding edges. It's staying on the field long enough to compound them.

How to Apply This to Your Betting

You don't need a spreadsheet.

Pick a bankroll amount. The money you've specifically allocated to betting, not your savings. Be honest about this number.

Apply the formula: Stake = (bankroll × 0.05) ÷ decimal odds

Round to a sensible whole dollar. Don't be a robot.

Revisit your bankroll size monthly. If it's grown, your stakes grow proportionally. If it's shrunk, your stakes shrink. This is the compounding mechanic that makes Unit Impact quietly powerful over a season.

At StatChecker, we're building the data tools that sit upstream of this, giving you the evidence to back the right bets in the first place. But even the best research is undermined by bad staking. The two work together.

The Takeaway

Kelly Criterion is the greatest staking framework ever devised. No model has surpassed it (Maclean, Thorp & Ziemba, 2010). But it requires accurate probability estimates that most punters simply don't have.

Unit Impact is the practical alternative. It's what Kelly looks like when you can't estimate precise probabilities but you understand that a $4.00 bet is a fundamentally different risk proposition to a $1.80 bet.

Flat staking ignores that. Unit Win punishes it. Unit Impact respects it.

Fifty-five thousand bets across professional tipsters confirms it. My own pro betting run confirmed it. The bankroll simulations confirm it.

Pick your number. Divide by odds. Round it. Bet it.

Evidence wins. So does discipline.

References

Barge-Gil, A. & García-Hiernaux, A. (2019). Staking plans in sports betting under unknown true probabilities of the event. MPRA Paper No. 92196, Universidad Complutense de Madrid.

Kelly, J. (1956). A new interpretation of information rate. Bell System Technical Journal, 35, 917–926.

Maclean, L.C., Thorp, E.O. & Ziemba, W.T. (2010). Long-term capital growth: the good and bad properties of the Kelly and fractional Kelly capital growth criteria. Quantitative Finance, 10(7), 681–687.

Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.

Thorp, E.O. (2008). The Kelly criterion in blackjack, sports betting, and the stock market. In Handbook of Asset and Liability Management. Elsevier.

Anthropic. (2026). Claude Sonnet 4.6 language model. Anthropic.