Stop Trading Fixed Lots: How Kelly Criterion Transforms Your F&O Risk Management
Fixed-lot sizing is silently destroying your algo PnL. Here's how Kelly Criterion and its fractional variants apply to Indian F&O trading.
Traders using fixed-lot sizing leave an estimated 30–40% of their edge on the table. Not because their strategy is wrong — because their sizing is.
Think about that for a second. You've spent months building and backtesting a strategy. Your win rate holds. Your risk-reward looks clean. But if you're still trading the same 1 lot on every signal regardless of confidence or account size, you're driving a Formula 1 car in second gear. The engine is there. The potential is there. The gear shift — position sizing — is what most traders never touch.
[!NOTE] TL;DR
- Fixed-lot trading ignores how strong or consistent your edge actually is, costing you compounding potential.
- Kelly Criterion is a mathematically grounded formula that tells you what fraction of your capital to risk per trade based on your win rate and reward-to-risk ratio.
- In Indian F&O, you should almost never use full Kelly — half-Kelly or quarter-Kelly is the practical standard.
- Recalculate your Kelly% every 30 trades or after a significant market regime change, not every day.
Before we go further: when did you last question your lot size? Was it set deliberately — or did you just pick 1 lot and never revisit it?
The Fixed-Lot Problem (With Numbers)
Let's make this concrete with BankNifty futures, currently trading around ₹52,000. One lot = 15 units. Margin requirement approximately ₹45,000–₹55,000 per lot (SEBI SPAN + exposure). Say your account is ₹5,00,000.
A trader running fixed 1-lot sizing risks roughly ₹7,500–₹10,000 per trade (a 150-point stop on BankNifty). That's 1.5–2% of capital — which actually sounds reasonable. But here's what fixed sizing does wrong: it never scales up when you're winning, and it never scales down when you're in a drawdown.
Look at this 10-trade sequence comparing fixed 1-lot sizing versus proportional Kelly-based sizing (half-Kelly, calibrated to the strategy stats we'll work through below). Starting capital: ₹5,00,000.
| Trade | Outcome | Fixed Lots | Fixed Capital (₹) | Kelly Lots | Kelly Capital (₹) |
|---|---|---|---|---|---|
| 1 | Win +₹8,000 | 1 | 5,08,000 | 1 | 5,08,000 |
| 2 | Win +₹8,000 | 1 | 5,16,000 | 1 | 5,16,480 |
| 3 | Loss −₹5,000 | 1 | 5,11,000 | 1 | 5,11,137 |
| 4 | Win +₹8,000 | 1 | 5,19,000 | 1 | 5,19,530 |
| 5 | Loss −₹5,000 | 1 | 5,14,000 | 1 | 5,14,085 |
| 6 | Loss −₹5,000 | 1 | 5,09,000 | 1 | 5,08,744 |
| 7 | Win +₹8,000 | 1 | 5,17,000 | 1 | 5,17,475 |
| 8 | Win +₹8,000 | 1 | 5,25,000 | 2 | 5,33,699 |
| 9 | Win +₹8,000 | 1 | 5,33,000 | 2 | 5,50,610 |
| 10 | Loss −₹5,000 | 1 | 5,28,000 | 2 | 5,40,198 |
Note: Kelly lot sizing rounds down to the nearest whole lot and steps up only when capital thresholds support it. Lot steps are illustrative for this example.
After 10 trades (7W/3L — a 70% win rate run), fixed sizing grows the account by ₹28,000. The Kelly approach — simply scaling up when the account grows enough to support another lot — produces ₹40,198 in the same sequence. That's a 43% improvement in P&L on an identical signal set.
Now run the reverse: a losing streak. Fixed sizing bleeds at the same rate regardless of your account balance. Kelly sizing naturally reduces exposure as capital falls, slowing the drawdown rate and preserving capital for recovery.
Here's the uncomfortable question: if your strategy produces 55% winning trades, why are you sizing your winners exactly the same as your losers?
Kelly Criterion Explained Without the Math Phobia
The Kelly Criterion was developed by John L. Kelly Jr. at Bell Labs in 1956. It was originally designed for signal transmission — then gamblers and investors realised it applied to any repeated bet with a known probability distribution.
The formula:
K% = W − (1 − W) / R
Where:
- K% = the fraction of your capital to risk on the next trade
- W = your historical win rate (as a decimal, e.g. 0.55 for 55%)
- R = your reward-to-risk ratio (average winner ÷ average loser)
Plain English: Kelly tells you to bet more when your edge is large and your wins are bigger than your losses, and less when your edge is thin.
Worked example — Nifty Futures strategy:
- Win rate: 55% (W = 0.55)
- Average winner: ₹8,000
- Average loser: ₹5,000
- Reward-to-risk ratio: ₹8,000 ÷ ₹5,000 = 1.6
Plugging in:
K% = 0.55 − (1 − 0.55) / 1.6
K% = 0.55 − 0.45 / 1.6
K% = 0.55 − 0.28125
K% = 0.26875
K% ≈ 26.9%
Full Kelly says risk 26.9% of your capital on this trade. On a ₹5,00,000 account, that's ₹1,34,375 per trade.
That number should alarm you. It should. Because it's supposed to.
[!WARNING] Common mistake: using full Kelly (1x) in real trading.
Full Kelly is the mathematically optimal fraction for long-run wealth maximisation — but it assumes your win rate and reward-to-risk ratio are known with perfect precision. They never are. In live F&O markets, both shift constantly. A 5% overestimate of your win rate, combined with a streak of losses, can produce drawdowns that wipe 50–60% of your account using full Kelly. Almost no professional systematic trader uses full Kelly. The standard in practice is half-Kelly (K% × 0.5) or quarter-Kelly (K% × 0.25).
Knowing the full Kelly number is useful — it anchors your thinking. But which fraction you actually deploy depends on how well you know your own edge. How confident are you in your strategy's statistics right now?
Fractional Kelly for Indian F&O
Here is how to think about Kelly fractions in the context of Indian derivatives:
| Kelly Fraction | Risk per Trade (₹5L account, K%=27%) | Best For | Drawdown Profile |
|---|---|---|---|
| Full Kelly (1×) | ₹1,34,500 (26.9% of capital) | Never in practice — theoretical baseline only | Severe; 50%+ drawdowns possible during bad runs |
| Half-Kelly (0.5×) | ₹67,250 (13.4% of capital) | Strategies with 200+ trade history and stable market regime | Moderate; drawdowns roughly 1/4 of full Kelly variance |
| Quarter-Kelly (0.25×) | ₹33,625 (6.7% of capital) | New strategies, volatile regimes, under-100-trade history | Conservative; slow compounding but robust to estimation error |
| Fixed 2% Rule | ₹10,000 flat (2% of capital) | Validation phase, new instruments, post-drawdown rebuilding | Minimal variance; underperforms in high-edge environments |
For most retail algo traders on NSE F&O, half-Kelly is the practical target once you have a robust backtest with 200+ trades. Quarter-Kelly is appropriate while you're still gathering live trade data or when the strategy is running in a higher-volatility regime (e.g., budget week, election results).
One important constraint specific to India: SEBI's margin framework sets minimum lot exposure irrespective of what Kelly recommends. You cannot trade 0.3 lots of BankNifty — you trade 1 lot or nothing. This means Kelly functions as a threshold system in F&O: you calculate the capital needed to justify one more lot, and you only add that lot when your account crosses the threshold.
[!TIP] Pro tip: recalculate your Kelly every 30 trades, not every day.
Daily recalculation introduces noise. If you had three bad days, your rolling win rate drops temporarily and Kelly would have you sizing near-zero just as the strategy recovers. Instead, maintain a rolling 30-trade window. Recalculate after every 30 completed trades, or after any structural regime shift — a new VIX regime, a major policy change, or a sustained period where your average win/loss profile changes by more than 15%.
Real Tradeoffs
Neither Kelly nor fixed lots is universally superior. Here is an honest comparison:
| Dimension | Fixed Lot | Kelly-Based Sizing |
|---|---|---|
| When it wins | When strategy stats are uncertain, during drawdown recovery, when account is too small to step up lot sizes | When strategy has a statistically validated edge (200+ trades), account is large enough to use fractional lots or scale across instruments |
| When it fails | Always underperforms Kelly in high-edge, well-validated strategies; fails to reduce exposure during drawdowns | Fails when win rate estimate is wrong (overfit backtests); lot-size granularity limits practical application in small accounts |
| Psychological load | Low — same action every trade | Medium — requires discipline to follow the formula even when it feels wrong |
| Implementation in algo | Trivial — hardcode 1 lot | Moderate — requires rolling trade log, Kelly calculation, margin check before order |
| Compounding effect | Linear growth | Geometric growth when edge is real; geometric decay when edge is overestimated |
Two questions to sit with:
Question 4: If you increased your lot size by 1 during your last winning streak and your last losing streak began immediately after — was that bad luck, or was your position sizing amplifying the pain?
Question 5: What would your account look like today if you had been cutting position size by 50% after every 5-trade losing streak, and doubling it back only after recovering the drawdown?
Choose Your Scenario
Scenario A: Your strategy has a stable, well-tested win rate (200+ trade history)
You have a live or backtested edge with at least 200 completed trades. Your win rate has been consistent across different market regimes — at least two to three distinct VIX environments. Your average winner and average loser are stable within a narrow band.
In this case, calculate your Kelly% using the formula above. Apply half-Kelly. Set a hard cap at 25% of capital per trade (no exceptions — see the DANGER callout below). Convert the rupee risk to lots using your current SEBI margin requirement. Run this sizing for 30 trades, then recalculate.
This is where Kelly earns its reputation. Your account compounds geometrically. Your losses hurt less because your exposure naturally contracts when your account dips. You are finally extracting the full mathematical value of your edge.
Scenario B: You are still validating your edge (under 100 live trades)
You have a promising strategy. Your backtest looks great. But you have fewer than 100 live trades in real market conditions.
Do not use Kelly here. Use a fixed 0.5–1% risk per trade. The reason is simple: backtested win rates are almost always optimistic. Curve-fitting, look-ahead bias, and the difference between backtest fills and real NSE order execution all erode the edge. Until you have live-market confirmation, your Kelly inputs are unreliable. Garbage in, garbage out — but with your actual capital.
Build the track record first. Run 100+ live trades with fixed micro-sizing. Then apply Kelly when you have numbers you actually trust.
5-Minute Position Sizing Framework
Use this decision flowchart before every trade. If you are running an algo, encode this logic in your pre-order risk check:
flowchart TD
A[New Trade Signal] --> B{Strategy has 100+ trade history?}
B -- No --> C[Use Fixed 0.5% Risk per Trade]
B -- Yes --> D[Calculate Win Rate & Avg RR]
D --> E[Apply Kelly Formula:\nK% = W - 1-W / R]
E --> F{K% > 25%?}
F -- Yes --> G[Cap at 25% — Full Kelly is Ruin]
F -- No --> H[Apply Half-Kelly: K% × 0.5]
G --> H
H --> I{SEBI Margin Available?}
I -- No --> J[Reduce to Available Margin]
I -- Yes --> K[Place Order with Calculated Lots]
C --> K
[!DANGER] Hard rule: never bet more than 2% of capital on any single F&O trade, regardless of what Kelly outputs.
Kelly assumes your probability estimates are accurate. In Indian F&O — where a single RBI surprise, a global risk-off event, or an options expiry move can gap through your stop — your estimates can be invalidated instantly. A 2% hard cap means even a catastrophic run of 10 consecutive full losses only draws down 18% of your account (compounding). That is survivable. Losses above 2% per trade start to compound into drawdowns that take years to recover from psychologically, not just financially. Set the cap in your algo's risk engine, not just in your head.
Mini-Exercise
Fill in this template for your current primary strategy. Do it now, not later.
My strategy win rate = [?]%
Average winner = ₹[?]
Average loser = ₹[?]
Kelly calculation:
W = [?] / 100
R = [avg winner] / [avg loser] = [?]
K% = W − (1 − W) / R = [?]%
Half-Kelly = [?]% × 0.5 = [?]%
My trading capital = ₹[?]
Risk per trade = ₹[?] (half-Kelly% × capital)
Instrument = [Nifty / BankNifty / MidcapNifty / Stock F&O]
Current SEBI margin = ₹[?] per lot
Max lots I can trade = [?] lots
If your risk per trade at half-Kelly exceeds your available SEBI margin for even 1 lot — you are either undercapitalised for this strategy at this sizing, or the strategy has a lower edge than your stats suggest. Both are important signals.
Two more questions before you close this tab:
Question 6: Have you ever entered more contracts than your pre-defined risk allowed because "this one felt different"? What happened after?
Question 7: If your algo fired 5 signals in the same session and you were running full Kelly on all of them simultaneously — what would your total portfolio exposure be? Have you thought about that?
Keep Learning
These posts build directly on the concepts above:
- Next: Momentum vs Mean-Reversion: Which Strategy Actually Works on Nifty? — understand which market regimes suit each approach before you calculate an edge worth sizing into.
- Then: Backtest NSE Futures Momentum in 50 Lines with vectorbt — generate the trade history your Kelly calculation needs, with Python code you can run today.
- Also: SEBI Algo Regulations for Retail India 2026: The Complete Guide — know exactly which SEBI rules govern margin, lot sizes, and automated order placement before you scale up.
Free Tool: Kelly Criterion Calculator for Indian F&O
I've put together a Kelly Criterion Calculator — Indian F&O Edition as a Google Sheets template. It includes:
- Auto-calculated Kelly%, half-Kelly%, and quarter-Kelly%
- Maximum lots based on your capital and current SEBI margin per instrument (pre-loaded for Nifty 50, BankNifty, MidcapNifty, FinNifty)
- A running trade log that updates your rolling win rate and R-ratio every 30 trades
- A colour-coded dashboard that flags when your Kelly% is being capped by the 2% hard rule
Get the Kelly Criterion Calculator — Free Google Sheets Template (link active — make a copy to your own Drive)
No email required. No paywall. Just use it.
Comment Below
What is your current risk per trade as a percentage of capital, and have you ever blown past it in the heat of the moment? Tell me the instrument and the lesson — the most honest answers get featured in a follow-up post on trading discipline.
This is not a rhetorical question. The most useful thing you can do after reading a post like this is write down one sentence: the time you ignored your sizing rules, the amount it cost, and what you would do differently. That sentence is worth more than the formula.
FAQ
Q1: Does Kelly Criterion apply to options buying, or only futures?
Yes, it applies to options buying too — with an important caveat. For options, your "average loser" is capped at the premium paid (maximum loss is defined), but your "average winner" can vary enormously between ITM near-expiry plays and far OTM lottery trades. Before applying Kelly to options, be precise about which type of options trade you are categorising. A 55% win rate on 200-point BankNifty spreads and a 55% win rate on ATM straddle buying are not the same strategy, even if the win rates look identical.
Q2: My backtest shows a 70% win rate. Can I use full Kelly?
Almost certainly not. Backtested win rates are systematically overestimated due to overfitting, look-ahead bias, and the difference between assumed fill prices and actual NSE execution (slippage on illiquid strikes, gap opens). As a heuristic: assume your live win rate will be 5–10 percentage points lower than your backtest. Run Kelly on the conservative estimate. If full Kelly still looks reasonable after that haircut, use half-Kelly anyway. The asymmetry of ruin — where a bad run destroys the account before a good run recovers it — always argues for caution.
Q3: How do I handle multiple simultaneous F&O positions in my Kelly calculation?
This is portfolio Kelly, and it is significantly more complex. The simple practical rule: if you are running N simultaneous uncorrelated strategies, divide your per-trade Kelly allocation by N. If the strategies are correlated (e.g., both long Nifty delta), treat them as one position for sizing purposes. SEBI's portfolio margin framework already recognises hedged positions — use that to your advantage, but do not double-count the risk reduction.
Q4: My strategy has a positive Kelly% but SEBI margin means I can only trade 1 lot. Is Kelly even relevant for small accounts?
Yes, in a directional sense. Kelly tells you whether your current edge justifies trading at all. If your Kelly% on half-Kelly works out to ₹8,000 risk per trade but the minimum BankNifty exposure is ₹45,000, your account is simply not yet capitalised for that instrument at that sizing. Either build the account further, switch to a lower-margin instrument (MidcapNifty, FinNifty, or stock futures with lower margin), or use fixed micro-sizing until your account grows to the Kelly threshold.
Q5: Is the Kelly Criterion legal under SEBI regulations?
Kelly Criterion is a position sizing methodology — it is a mathematical framework applied to your own capital allocation decisions. It is not a trading strategy, a signal system, or an advisory service. SEBI regulations govern order types, algo registration, and margin requirements, not how you calculate your internal risk per trade. Using Kelly is no different from deciding to risk 2% per trade — it is a personal risk management decision. This post is for educational purposes only and does not constitute SEBI-regulated investment advice.
Do This Next
- Write down your strategy's win rate and average winner/loser from your last 30–50 trades (use your broker's trade history export if needed).
- Plug those numbers into the Kelly formula:
K% = W − (1 − W) / R. Calculate half-Kelly. - Check whether your current lot size corresponds to a Kelly%, half-Kelly%, or fixed percentage of capital. You may be surprised at the mismatch.
- Make a copy of the Kelly Criterion Calculator spreadsheet and enter your strategy's live trade data into the running log.
- Set a calendar reminder for 30 trades from today to recalculate your Kelly inputs and update your sizing accordingly.
- Add a pre-order risk check to your algo (or a manual checklist for discretionary trades) that enforces a hard 2% capital cap regardless of Kelly output.
- Read the SEBI Algo Regulations 2026 guide to confirm your position sizing and order flow automation is compliant before you scale.
This post is for educational purposes only. It does not constitute financial or investment advice, and is not a SEBI-registered research or advisory service. All examples use hypothetical figures. Past performance of any strategy or sizing method does not guarantee future results. Trading F&O involves significant risk of loss.