A retail trader in Pune spent three weekends building a moving average crossover bot. The backtest showed an 18% return over six months with a clean equity curve. He went live with real money in the first week of the new SEBI algo framework. Within three weeks, he had lost 40% of his capital, and the bot had technically followed its rules the whole time.
This happens more often than most beginners expect, and it rarely comes down to a bad strategy. It usually comes down to a handful of avoidable mistakes that never show up in a backtest. Here are the five that catch retail algo traders in India most often.
1. Treating the backtest as a promise, not an estimate
A backtest tells you how a strategy would have performed on historical data, assuming perfect fills, no slippage, and no emotional interference. None of those assumptions hold in live markets. If your backtest doesn’t account for spread, brokerage, and STT, the real number will be lower, sometimes significantly. Before going live, run the same strategy on out of sample data it has never seen, and stress test it against a period with a sharp drawdown, like early 2020 or late 2022.
2. Ignoring slippage and execution delay
Retail algo systems built on broker APIs don’t execute instantly. There’s a gap between the signal firing and the order actually filling, and in fast moving stocks that gap costs money. A strategy that looks profitable at a theoretical entry price can turn unprofitable once you factor in even a 0.1% average slippage per trade. Log your actual fill prices against your intended entry prices for the first month and compare them honestly. [Also, read Best Trading API in India]
3. Sizing positions based on conviction, not capital
Once a bot has had a few winning trades, it’s tempting to increase position size because “the system is working.” This is how a strategy with a genuinely good win rate still blows up an account, because a string of losses at oversized positions does more damage than the same string at consistent sizing. Fixed fractional position sizing, where you risk a set percentage of capital per trade, protects you from this even when the strategy itself is sound. Zerodha’s Varsity education portal has a solid module on risk management and trading psychology that’s worth going through before scaling up size.
4. Not registering under SEBI’s new algo framework
Since SEBI’s October 2025 rollout of the retail algo trading framework, algos that cross a certain order-per-second threshold need to be registered through your broker, tagged with a unique exchange identifier for audit purposes. Traders who built and ran self-developed algos before the framework sometimes assume old habits still apply. They don’t. We covered the specifics of what changed in our guide to the SEBI rules for retail algo trading, and it’s worth reading the original SEBI circular directly if you’re running anything beyond basic order automation.
5. Skipping a kill switch
Every live algo needs a manual override that can flatten all positions and stop new orders instantly, independent of the strategy logic itself. This sounds obvious until an API disconnects mid session, or a strategy hits an edge case the backtest never covered, and there’s no way to intervene except pulling the internet cable. A kill switch is a few lines of code. Not having one is how a small bug becomes a large loss.
None of these mistakes are about strategy quality. They’re about the gap between a clean backtest and a messy live market, and closing that gap is most of what separates retail traders who last from ones who don’t. If you’re just getting started, our tips for retail traders guide is a good place to build the fundamentals before any of this becomes relevant.
