Custom backtesting vs no-code platforms: where the drop-downs run out

Stockmock, AlgoTest, Streak and the other Indian strategy builders are genuinely useful, and for a large share of traders they are the correct answer. This is not an argument against them. It is an attempt to describe precisely where their ceiling sits — because if your strategy is below it, you should not be paying anyone to code anything.

What the no-code platforms do well

Credit where it's due. A good strategy builder gives you:

• Speed. An idea to a result in ten minutes, with no specification, no developer and no waiting.
• Clean data you didn't have to source. Options chains, expiry handling and contract rollovers already dealt with — which is genuinely hard and genuinely tedious.
• Cheap iteration. Testing thirty variants of the same straddle costs you an afternoon rather than thirty invoices.
• A sane cost model. Most of them apply brokerage and taxes by default, which puts them ahead of the average DIY spreadsheet.

If your strategy is "short a Bank Nifty straddle at 09:20, stop at 30% of premium, exit at 15:10", a builder will test it faster and cheaper than any human, and hiring someone would be a waste of your money. Use the tool.

The ceiling is not a flaw — it's the design

A no-code builder works by offering a fixed vocabulary: a list of entry types, a list of exit conditions, a list of instruments. That constraint is exactly what makes it fast and reliable. It also means the platform can test every strategy expressible in that vocabulary, and no others.

The moment your logic needs a word the vocabulary doesn't contain, you have three options: distort your strategy until it fits the drop-downs, abandon the idea, or have it coded. Most traders quietly do the first, then wonder why the live results don't match the plan they had in their head.

The one-question test: can you express your rule exactly, with no "close enough", using only the options the builder gives you? If yes, use the builder. If you found yourself substituting a similar-but-not-identical condition, you have already left the tool's range — and you are now testing a different strategy from the one you want to trade.

The five walls people actually hit

1. A custom indicator or your own formula

Builders offer a standard library: moving averages, RSI, supertrend, VWAP. If your signal uses a proprietary combination, a custom volatility measure, a ratio between two instruments, or anything you derived yourself, it simply isn't in the list. This is the most common wall by a wide margin.

2. Conditional or state-dependent logic

"Enter the second leg only if the first is in profit by 20 points." "Skip today entirely if yesterday closed beyond 1.5 ATR." "Trail the stop only after 11:00, and only if the trend filter is still valid." These are conditions about other conditions, and most builders are flat rule lists with no memory of state. You cannot express a dependency in a drop-down.

3. Data the platform doesn't carry

India VIX as a filter, an FII/DII flow input, a global index as a bias signal, corporate-action-adjusted equity history, a custom universe of forty stocks screened by your own criteria. If the platform doesn't already carry the series, no amount of configuration will produce it.

4. Portfolio-level thinking

Builders test one strategy on one instrument. Real capital allocation asks harder questions: how do three strategies behave together? Are their drawdowns correlated, and do they arrive on the same days? What is the right allocation between them? What does total exposure look like when all three fire at once? A per-strategy result cannot answer a portfolio question.

5. Robustness testing, not just a result

Most builders show you what happened. The more important question is whether what happened was luck. That needs a parameter sensitivity sweep, walk-forward validation and often a Monte Carlo study on trade-order randomisation. A single equity curve, however pretty, tells you nothing about how fragile it is — and fragility is what kills accounts.

What custom actually costs, honestly

The comparison people make is "₹1,000 a month subscription versus ₹5,000 one-off", which frames custom work as the expensive option. That's the wrong frame in both directions, so here is a fairer one:

Note the last row, because it's the one people undervalue. A subscription rents you the answer. A custom backtest hands you the machine that produced it — so re-running with different parameters next quarter costs nothing, and the same code becomes the live strategy rather than a separate reimplementation that may or may not match.

The sequence that wastes the least money

For most traders the right order is not "pick one". It's:

1. Prototype in the builder — as close to your idea as the drop-downs allow. If the approximation already loses money, stop. You just saved yourself the whole exercise for the price of an afternoon.
2. Note exactly what you had to compromise — the filter you left out, the condition you simplified. That list is your specification.
3. Get the uncompromised version tested properly, with sensitivity and walk-forward, only if the approximation showed something worth pursuing.
4. Automate it — in Pine Script or MQL5 — only once the honest version survives testing.

Skipping step one is how people spend money testing ideas that a free tool would have killed in ten minutes. Skipping step three is how people go live with a strategy they never actually validated.

When you should not hire anyone

Bluntly, because it matters more than the sales pitch:

• Your strategy fits the builder. Then use it. Paying for custom work is pure waste.
• You haven't defined your rules yet. No developer can test "I buy when it looks strong". Define it first — the builder is a decent forcing function for that.
• You want to test forty variants. That's what subscriptions are for. Come back when you have one or two candidates worth a rigorous look.
• You're hoping the backtest says yes. If a negative result would upset you rather than inform you, you're not buying research — and this is the wrong purchase.

The underlying point

No-code platforms democratised backtesting, and that was a real good. What they cannot do is test the strategy that only exists in your head — the one with the unusual filter, the conditional leg and the custom indicator you built over two years of screen time. That strategy is also, not coincidentally, the one most likely to have an edge, precisely because it isn't in everyone else's drop-down menu.

The tool tests common strategies well. Uncommon strategies are where the edge tends to live, and testing those is a coding problem.