MyAlgoGuru Help
Education Glossary Open the app
Chapter 5 of 5

Why the curve and the statement disagree

Four layers sit between a gross backtest number and money in your account. Most reconciliation arguments are really an argument about which layer you forgot.

A trader runs a strategy for a month. The platform's report says ₹46,000. The broker's statement says ₹31,000. Somebody is wrong.

Usually nobody is. The two numbers are measuring different things, and the gap is a stack of four layers, each of which is computed somewhere else. Learning to walk down the stack turns a frustrating discrepancy into a two-minute reconciliation.

100 gross what the rules earned −18 charges brokerage, STT, exchange, GST −9 slippage the fill, not the quote −5 account platform fees, manager share 68 net before income tax — not modelled
Four layers, each computed in a different place. Gross is the rules alone. Charges come from the cost model and the segment. Slippage is an assumption about your own execution. Account charges are levied on the account rather than the trade, so they never appear per-trade anywhere. Income tax is not a trading cost at all and no cost model here touches it. The proportions drawn are illustrative — on a long-horizon strategy the red blocks are slivers, and on a scalping strategy they can be most of the picture.

Walking down the stack

Layer one: is the gross number gross? Check whether costs were applied in the run you are comparing. In this platform's backtester transaction costs are off by default, and the bill is estimated and shown beside the result even when it is not deducted. Many reconciliation arguments end here, with one party reading a gross figure and the other a net one.

Layer two: what slippage did the run assume? A run at zero slippage assumes you always got the quoted price. If you did not, the gap belongs here, and the only way to know its size is the measurement from the previous chapter.

Layer three: what does the account charge, as opposed to the trade? Platform fees, annual maintenance, and — on a managed account — the manager's share of profits. These are levied per account per period, so they are invisible in any per-trade view. They are also where the single most expensive arithmetic error in this whole module lives.

Layer four: income tax. Not a trading cost, not in any model here, and large. How your results are classified and what you owe on them is a question for your accountant, and it materially changes what you keep. Nothing on any screen in this platform is an after-tax number.

The hundred-times error

Account-level charges are usually configured as a percentage. That innocent word is responsible for an entire class of bug, and it is worth understanding because you will meet it in your own spreadsheets long before you meet it anywhere else.

A field labelled "fee percent" holding the value 2 means two percent. To apply it you must divide by a hundred: profit × 2 ÷ 100. A field labelled "fee rate" holding 0.02 means the same thing, and to apply it you multiply directly: profit × 0.02.

Mix them up and you are out by a factor of a hundred. On a ₹3,00,000 profit, a 2% manager fee is ₹6,000; the same number applied as a rate is ₹6,00,000 — twice the profit it is charged on. The failure is loud in that direction, and it is silent in the other: a rate of 0.02 treated as a percent gives ₹60, which looks like a plausible small fee and quietly understates the charge forever.

The discipline that prevents it is to name the unit in the field, not in a comment, and to convert at exactly one place in the code. If a value's name does not say whether it is a percent or a fraction, the next person to read it has to guess, and eventually one of them guesses wrong.

A sum of some legs is not the basket

One more trap, and it is the reason a multi-leg P&L figure sometimes carries a warning rather than a number.

A four-leg basket's profit is the sum of its four legs. If one leg's price is unavailable — an untraded strike, a stale feed, a contract the instrument master does not carry — then the sum of the other three is not the basket's profit. It is a smaller, unrelated number that looks exactly like a profit figure.

The honest behaviour is to show the partial sum and say that it is partial, so a reader never mistakes three legs for four. A screen that quietly sums what it has will, on exactly the days when something is wrong with the data, report a number that is confidently incorrect. If you are building your own tracking in a spreadsheet, build the same check: count the legs you priced, compare it with the legs you hold, and refuse to report a total when they disagree.

A reconciliation worth doing monthly

Illustrative

Platform report: ₹46,000, costs not applied, slippage zero.

Step Working Running
Gross, as reported ₹46,000
Transaction costs 142 round trips × ₹78 ₹35,324
Slippage measured at ₹0.11 per unit, 142 trips × 75 units ₹34,152
Platform fee flat monthly ₹33,652
Expected net ₹33,652

Broker statement: ₹31,000.

A gap of ₹2,652 remains — about 8%. That is small enough to be real slippage running worse than measured, and large enough to be worth one look. The useful next step is not to re-run the backtest; it is to pull a handful of individual fills and compare each one against the quote at signal time. Either the slippage assumption is low, in which case update it and every future report improves, or a few specific trades filled badly, in which case you have found something about your execution rather than about your strategy.

What you must not do is adjust the backtest until it matches. That is fitting the test to the statement, and it destroys the only independent check you have.

Do this once a month and two useful things happen. Your slippage assumption becomes an actual measurement, so every subsequent report is closer to the truth. And a gap that suddenly widens becomes a signal — something changed in your execution, your broker's fills or your plumbing, and you found it in a reconciliation rather than in a drawdown.

What to take from this chapter

  • Gross and net differ by four layers: transaction charges, slippage, account-level charges, and income tax. Most disagreements are about a forgotten layer, not an error.
  • Check first whether costs were applied at all. In this platform's backtester they are off by default, with the bill shown beside the gross result.
  • Account charges are levied per account, not per trade, so they are invisible in any per-trade view. Add them by hand.
  • Never let a percent and a fraction share a name. One conversion, one place, unit in the field name — or you will be out by a hundred times, loudly in one direction and silently in the other.
  • A sum of three legs out of four is not a basket's P&L. Report it as partial or not at all.
  • Reconcile monthly, and fix your slippage assumption from what you find. Never adjust the backtest to match the statement.