I have the spreadsheet open in front of me as I write this. It is a trade log, 412 rows, EUR/USD, dated across a single quarter. Column F is labelled "Kelly %". Row after row, it reads 0.31, 0.34, 0.29 — a man betting roughly a third of his account on every signal because a formula told him to. The account balance column starts at ₹8,40,000 and ends, eleven weeks later, at ₹1,12,000.
That is what the practical application of the Kelly criterion looks like when nobody warns you first. So let me warn you. The maths is sound. The way most retail traders use it is not. Below are the red flags I have watched sink accounts — and the books that taught me to spot them.
TL;DR
- Full Kelly sizing survives the maths, not the drawdown.
- Your "edge" is a guess dressed as a number.
- Spreads and SEBI rules quietly delete the formula's assumptions.
Red Flag #1: You Are Sizing at Full Kelly
Here is what it looks like. You estimate a 55% win rate, an even payoff, and the formula spits out a fraction — say, 10% of equity per trade. You key it in and feel scientific.
Here is why it matters. Kelly maximises long-run growth, yes, but it does so at the cost of a stomach-churning ride. At full Kelly, a 50% drawdown is not a tail event — it is an ordinary Tuesday. The original 1956 work by John Kelly Jr. at Bell Labs assumed a known, fixed edge and infinite divisibility. Forex gives you neither.
Ed Thorp, the man who actually traded Kelly for decades, never ran full Kelly. He ran a fraction of it. That single insight — half-Kelly, quarter-Kelly — is the most useful thing in the entire literature, and it is the first thing the Telegram groups skip.
Red Flag #2: Your Edge Is a Guess Wearing a Number
What it looks like: you pull 90 days of backtest results, calculate a win rate to two decimals, and feed it in as gospel.
Why it matters: Kelly is brutally sensitive to the inputs. Overstate your win rate by five percentage points and the formula tells you to bet roughly twice as much as you should. Get the edge wrong and the same equation that compounds your money also compounds your ruin.
I rate Thorp's *A Man for All Markets* the single most honest book here precisely because he keeps repeating that estimating the edge is harder than applying the formula. The formula is one line of arithmetic. The edge is a lifetime of work. Anyone selling you a "Kelly calculator" without making you sweat over the win-rate input is selling you a faster way to be wrong.
Red Flag #3: You Treat Forex Like a Coin Flip
What it looks like: you plug in win probability and a fixed reward-to-risk ratio, as though every EUR/USD trade resolves to a clean win or a clean loss.
Why it matters: the original Kelly model was built for binary bets — the wire-tap, the horse race, the blackjack hand. Forex outcomes are continuous. Your stop gets slipped. Your take-profit gets partially filled. A "1:2 trade" becomes a 1:1.3 trade because price gapped over the weekend.
The book that wasted my time here was every generic "money management" PDF that ports the gambling formula straight onto a chart without adjusting for continuous outcomes. The serious treatment — the continuous Kelly used by professional desks — needs the variance of returns, not a win/loss tally. If your method does not even mention variance, it is the wrong tool.
Red Flag #4: The Spread Is Not in Your Edge Calculation
What it looks like: you compute your edge on raw price moves and ignore what the broker takes on the way in and out.
Why it matters: your real edge is the after-cost edge, and costs are not trivial. Take Exness, whose standard EUR/USD spread sits around 1.0 pip per their published schedule. On a standard 100k lot that is 1.0 pip × $10/pip × USD/INR 83.42 = ₹834.20 per round trip, before you have made a single rupee. Trade twenty round trips a week and you are handing over ₹16,684 weekly just to participate.
Kelly assumes a positive edge. If your edge before costs is thin, the spread can flip it negative — and the formula, fed a phantom positive edge, will happily size you into a guaranteed grind toward zero.
Red Flag #5: You Confuse the Kelly Fraction With Broker Leverage
What it looks like: the formula says risk 8% of equity, so you reach for an account offering 1:2000 leverage to "make it efficient."
Why it matters: the Kelly fraction is about how much of your capital is genuinely at risk — your stop distance times your position size. Leverage is just the financing that lets you hold the position. Exness advertises up to 1:2000; that number has nothing to do with what Kelly is telling you to risk.
I have watched traders read "8% Kelly" and then deploy 1:500 because the broker allowed it, turning a disciplined sizing rule into a margin-call machine. The leverage figure is a feature of the account. The risk fraction is a decision about survival. Do not let the broker's marketing number quietly overwrite the maths.
Red Flag #6: Your Edge Expired and You Did Not Notice
What it looks like: you computed your win rate last December and have been sizing off it ever since.
Why it matters: Kelly assumes a stationary edge — the odds do not change between bets. Forex regimes change constantly. A breakout system that printed money in a trending quarter becomes a chop-fed loser the moment volatility compresses. The win rate you measured is a photograph of a market that no longer exists.
The book that changed how I trade here was Nassim Taleb's *Fooled by Randomness* — not a Kelly book at all, but the antidote to it. It teaches you that a clean historical track record is often just survivorship and luck wearing a suit. Re-estimate your edge on a rolling basis, or accept that you are sizing off a fossil.
Red Flag #7: You Are Running Correlated Pairs as Separate Bets
What it looks like: you size EUR/USD, GBP/USD, and AUD/USD each at quarter-Kelly and tell yourself you are diversified.
Why it matters: those three pairs share a dollar leg. When the dollar moves, they move together. Three "independent" quarter-Kelly bets can behave like one three-quarter-Kelly bet on the US dollar — exactly the over-betting Kelly was supposed to protect you from.
This is where most retail position-sizing advice falls apart. The single-asset formula assumes each bet is independent. The multi-asset Kelly requires the covariance matrix, which almost nobody computes. The practical fix is humbler than the maths: treat correlated positions as one position, and size the cluster, not the tickers. Pretending otherwise is how a "diversified" book takes a single, concentrated punch.
Red Flag #8: You Are an Indian Resident Trading Offshore FX Pairs
What it looks like: you open an offshore account, fund it, and start applying Kelly to EUR/USD as though the only constraint is your own discipline.
Why it matters: for a resident Indian, it is not. Under RBI's Liberalised Remittance Scheme and SEBI's framework, residents may legally trade currency derivatives only on recognised Indian exchanges — and only INR pairs such as USD/INR, EUR/INR, GBP/INR and JPY/INR. Margin forex on offshore EUR/USD through brokers like Exness or XM sits in a grey-to-prohibited zone, and remittance for it is not a sanctioned LRS purpose.
The cleaner compliance path is a domestic SEBI broker. Bajaj Finserv Securities opens a digital demat in about five minutes with PAN, Aadhaar and a linked bank account, zero AMC in year one. No Kelly fraction is worth an FEMA problem.
The Verdict
The Kelly criterion is not a scam, and I want to be clear about that. It is one of the most important ideas in the mathematics of betting, and a fractional version of it — half or quarter Kelly — is a genuinely sane way to think about position size once you have an edge you trust.
But "once you have an edge you trust" is doing all the work in that sentence. For most retail forex traders the honest edge is unknown, the costs are underestimated, the positions are correlated, and the legal footing — for an Indian resident on offshore pairs — is shaky. If you are starting out, open a compliant domestic demat with Bajaj Finserv Securities, trade INR pairs, log every trade, and only reach for Kelly after you have a year of real, after-cost data. The formula is the last 5% of the work. The other 95% is the part nobody puts in a Telegram message.
FAQ
What fraction of Kelly is actually safe for forex?
There is no universally safe number, but practitioners who trade Kelly with real money almost never run full Kelly. Half-Kelly captures roughly three-quarters of the long-run growth with materially smaller drawdowns; quarter-Kelly is calmer still. The honest answer is that your fraction should shrink as your confidence in the edge estimate shrinks — and for most retail traders, that confidence should be low, which argues for a small fraction or none at all until the data is in.
How do I include the spread in my Kelly edge?
Compute your edge on net outcomes, not raw price moves. Subtract the round-trip cost from every trade before measuring win rate and average payoff. On Exness's standard EUR/USD spread of about 1.0 pip, that is roughly ₹834.20 per 100k lot round trip at USD/INR 83.42. If subtracting realistic costs turns your historical edge negative or marginal, Kelly will size you toward slow ruin — which is exactly the warning you want before funding the account.
Can a resident Indian legally use Kelly on EUR/USD?
The sizing maths is legal anywhere; the underlying trade may not be. Under RBI's Liberalised Remittance Scheme and SEBI rules, residents may trade currency derivatives only on recognised Indian exchanges and only in approved INR pairs. Margin forex on offshore EUR/USD through brokers such as Exness or XM is not a sanctioned LRS purpose. Apply Kelly to USD/INR on a domestic SEBI broker instead — the formula does not care which pair you feed it.
Why is full Kelly considered dangerous if it is mathematically optimal?
It is optimal for long-run growth only under perfect inputs — a known, fixed edge and infinitely divisible bets. Real forex violates both. Full Kelly's "optimal" path runs through drawdowns of 40-50% as routine events, and a single overstated win rate makes it over-bet by a wide margin. Most traders cannot psychologically hold a position through that ride, so they abandon the system at the worst moment. Fractional Kelly trades a little theoretical growth for a survivable experience.
Which book should I read first on this topic?
Start with Ed Thorp's *A Man for All Markets*. He actually traded Kelly through decades of real markets, and he is relentlessly honest that estimating your edge is far harder than applying the formula. Pair it with Taleb's *Fooled by Randomness* as a corrective against trusting a clean backtest. Skip the generic "money management" PDFs that port the gambling formula onto charts without adjusting for continuous outcomes, variance, or trading costs — they teach the equation and hide the assumptions.
That spreadsheet I opened at the start: row 412, Kelly % column, reads 0.29. The balance column beside it reads ₹1,12,000, down from ₹8,40,000. Eleven weeks. Full Kelly, on an edge that was never measured after costs. That is the number. It is sitting in front of me.