Home / Data Quality / Bad Ticks & Outliers
DataIntermediate

Bad Ticks & Outliers

Bad ticks are erroneous data points — fat-finger prints, misfeeds, decimal errors, stale or off-market prices — that do not reflect a real tradeable market, and a single one can spike an indicator, trigger a stop or fire a breakout, creating a phantom signal that the backtest treats as genuine.

Quick answer: Bad ticks are erroneous data points — fat-finger prints, misfeeds, decimal errors, stale or off-market prices — that do not reflect a real tradeable market, and a single one can spike an indicator, trigger a stop or fire a breakout, creating a phantom signal that the backtest treats as genuine.

In simple words

A bad tick is a price in your data that is simply wrong: a typo, a glitch, a trade far away from the real market, or a spike that lasted a fraction of a second and could never have been traded at scale. The problem is that a backtest cannot tell a real move from a bad tick unless you clean it. One erroneous spike can cross a stop or set off a signal, generating a trade that should never have existed and distorting the whole result.

Purpose

Detecting and filtering bad ticks exists because raw market data is never perfectly clean, and an unfiltered outlier is indistinguishable to a backtest from a real opportunity, so it manufactures trades out of noise.

Professional explanation

Where bad ticks come from

Erroneous data has many sources: a trader entering the wrong price or quantity (a fat-finger error), a decimal or scaling mistake in a feed, a stale quote that did not update, an off-market print from a block or auction trade that does not reflect the continuous market, or a transmission glitch that garbles a value. Some are wildly obvious — a price ten times the real level — while others are just plausible enough to slip through. The defining feature is that no rational trader could have transacted meaningful size at that price at that moment, so it should not drive a trading decision.

How one bad tick becomes a phantom signal

A backtest computes signals mechanically, so it has no common sense about a price that is obviously wrong. A single spike high can register as a breakout and trigger an entry; a spike low can pierce a protective stop and book an exit that never happened; an outlier can inflate a volatility estimate or distort a moving average for the length of its window. Because the bad tick is often a large deviation, its effect on any threshold-based rule is disproportionate. The trade it creates is entirely fictitious, yet it enters the backtest's statistics as if it were real, biasing win rate, drawdown and expectancy.

Detection methods

Outlier detection blends simple range checks with statistical tests. A basic filter rejects prices outside a plausible band relative to the previous close, or moves larger than a multiple of recent volatility — for example a jump beyond several times the recent standard deviation of returns that immediately reverses. More refined approaches flag a tick that is inconsistent with the surrounding quotes, that has no corresponding volume, or that carries a non-standard trade condition marking it as off-market. The art is calibration: too loose and bad ticks survive, too tight and real gap moves — an earnings shock, a circuit move — are wrongly deleted as outliers.

Filtering without destroying real information

The danger of aggressive cleaning is that genuine large moves look exactly like bad ticks. A stock that legitimately gaps twenty percent on results, or an index that moves sharply on a policy announcement, must not be smoothed away, because those are precisely the events a robust strategy has to survive. Good practice distinguishes a transient spike that reverses within a tick or two from a sustained move to a new level, and it prefers to flag and review rather than silently overwrite. Corrections should be logged and, where possible, replaced with a defensible value such as the surrounding quote midpoint rather than an interpolation that invents a path.

Why it corrupts a backtest specifically

In live trading a human or a sanity check would usually ignore an obviously erroneous print, and exchange rules often cancel clearly mistaken trades. A backtest has none of that judgement, so an unfiltered bad tick is more dangerous in simulation than it would be in reality. This asymmetry means the backtest can show trades that could not have occurred live, systematically distorting results in a direction that depends on where the spikes fall. Cleaning bad ticks is therefore not cosmetic; it is what keeps the simulated market connected to the tradeable one.

Bad tick vs Real large move

AspectBad tickReal large move
DurationOne or few ticks, then reversesSustained at the new level
VolumeOften none or trivialReal volume transacts
Trade conditionMay be off-market or flaggedNormal continuous trade
Correct actionFlag, filter or correctKeep — it is genuine risk
Effect if keptPhantom signal, false stopLegitimate part of the test

Practical example

Illustrative example (Indian market)

You backtest a Bank Nifty breakout system on intraday data with capital of Rs 5,00,000, entering long on a break above the prior swing high of 48,200. One tick in your feed misprints as 49,150 for a single instant before reverting to 48,190, an obvious fat-finger or feed glitch with almost no volume. Your unfiltered backtest reads that spike as a decisive breakout, enters long at around 49,150, and then watches price sit near 48,190 — an immediate paper loss of nearly 960 points that never corresponded to a real trade. Filtering the spike, because it lasted one tick, carried no volume and reversed instantly, removes the phantom entry entirely. Across a long intraday history, a handful of such uncaught spikes can dominate the strategy's apparent drawdown or, if they fall favourably, its apparent gains.

Illiquid NSE stocks and far out-of-the-money options are where bad ticks cluster, because a single stray order can print far from the real market and no offsetting volume corrects it quickly. A backtest on such instruments that does not filter outliers will generate phantom breakouts and stop-outs, and because the affected names are thin, those fictitious fills would also have been impossible to execute at size in reality.

Limitations

  • Genuine gap moves on news look identical to bad ticks, so filtering can delete real risk
  • Outlier thresholds require calibration; too loose keeps errors, too tight removes real events
  • Illiquid instruments have both more bad ticks and thinner data to validate against
  • Silent overwriting of a spike can hide a data-source problem that should be investigated
  • A filter tuned on one instrument or regime may misclassify moves in another

Common mistakes

  • Running a backtest on raw feed data without any outlier detection at all
  • Filtering so aggressively that legitimate earnings gaps or circuit moves are smoothed away
  • Replacing a bad tick by interpolation, which invents a price path that never occurred
  • Ignoring volume and trade-condition flags that would distinguish a spike from a real move
  • Assuming a clean vendor feed needs no checking, when even good feeds carry occasional errors
  • Letting a single spike dominate volatility or drawdown statistics without investigating it

Professional usage

Professional data teams run outlier detection as a standard pipeline stage, combining range and volatility-based checks with volume and trade-condition filters, and they flag suspect ticks for review rather than silently deleting them. They deliberately preserve genuine large moves by distinguishing transient reversing spikes from sustained repricings, and they log every correction so the cleaning is reproducible and auditable. They also recognise that instruments cluster errors by liquidity, tightening scrutiny on thin names, and they cross-check a suspicious print against an independent feed before trusting or removing it.

Key takeaways

  • A bad tick is an erroneous price at which no meaningful trade could really have occurred
  • A single bad tick can trigger a phantom breakout, a false stop or a distorted indicator
  • Detect outliers with range, volatility, volume and trade-condition checks, calibrated carefully
  • Do not smooth away genuine news gaps or circuit moves — those are real risk the strategy must face
  • Flag and log corrections rather than silently overwriting, and scrutinise illiquid names harder

Frequently asked questions

What is a bad tick?
A bad tick is an erroneous data point — a fat-finger price, a decimal or feed error, a stale quote or an off-market print — that does not reflect a real tradeable market. The defining feature is that no rational trader could have transacted meaningful size there at that moment.
How does a bad tick affect a backtest?
A backtest applies its rules mechanically, so a bad tick can register as a breakout, pierce a stop, or distort an indicator. The trade it creates is fictitious but enters the statistics as if real, biasing win rate, drawdown and expectancy.
What is a phantom signal?
A phantom signal is a trade triggered by erroneous data rather than a genuine market move. A single spike high that reads as a breakout, or a spike low that trips a stop, produces a trade that could never have happened live.
Where do bad ticks come from?
From fat-finger order-entry errors, decimal or scaling mistakes in a feed, stale quotes, off-market block or auction prints, and transmission glitches. Some are obviously wrong, while others are just plausible enough to pass unnoticed.
How do I detect bad ticks?
Combine range checks against the previous close with volatility-based tests that flag moves beyond several times recent return standard deviation that immediately reverse, and add volume and trade-condition filters. Calibration matters, because thresholds too tight delete real moves.
How do I tell a bad tick from a real gap?
A bad tick is usually a single or few ticks that reverse instantly, often with no volume and sometimes an off-market flag. A real move is sustained at the new level with genuine volume, so duration, volume and trade condition distinguish them.
Should I delete or correct a bad tick?
Prefer to flag and review, then correct with a defensible value such as the surrounding quote midpoint rather than interpolating a path. Log every correction so the cleaning is reproducible, and avoid silently overwriting, which can hide a feed problem.
Can filtering bad ticks remove real information?
Yes, and that is the main risk. Legitimate news gaps, earnings shocks and circuit moves look like outliers, so over-aggressive filtering can smooth away exactly the large moves a robust strategy must survive.
Why are illiquid instruments more prone to bad ticks?
Because a single stray order can print far from the real market and there is little offsetting volume to correct it quickly. Thin NSE stocks and far out-of-the-money options therefore cluster both bad ticks and impossible fills.
Are bad ticks worse in a backtest than in live trading?
Often yes. Live, a human or a sanity check would ignore an obviously wrong print and exchanges may cancel mistaken trades, but a backtest has no such judgement, so an unfiltered bad tick creates trades that could not have occurred.
Do bad ticks affect daily data or only ticks?
Mostly intraday and tick data, but daily bars can carry a bad high or low from a single erroneous print during the session. A spurious daily extreme can still trip a stop or breakout, so daily data needs sanity checks too.
How does volume help detect bad ticks?
A genuine move transacts real volume, whereas many bad ticks print with little or no volume because no one actually traded there. Cross-referencing a suspicious price against its volume is a quick and effective screen.
What is an off-market print?
An off-market print is a reported trade from a block, auction or negotiated deal that does not reflect the continuous order-driven market. It can appear far from the prevailing price and should be excluded from a series used for signals.
How do bad ticks relate to overall data quality?
Bad ticks are one facet of data quality alongside missing values, corporate actions and adjustment. All of them concern whether the data faithfully represents a tradeable market, and unhandled bad ticks specifically inject false trades into an otherwise sound backtest.

Voice search & related questions

Natural-language questions people ask about Bad Ticks & Outliers.

What is a bad tick?
It is a wrong price in your data — a typo, a glitch, or a spike that lasted an instant. It looks like a real price to a backtest, but you could never actually have traded at it.
Why is one bad tick a big deal?
Because your backtest follows rules blindly. A single spike can look like a breakout or hit your stop, creating a trade that never really happened and messing up your results.
How do I spot a bad tick?
Look for a price that jumps far from the market for just an instant and then snaps back, usually with little or no volume. A real move stays at the new level and has real trading behind it.
Can cleaning bad ticks hurt my data?
Yes, if you overdo it. A genuine news gap can look like an error, so cleaning too hard can erase real risk your strategy needs to face. Flag and check rather than blindly delete.
Are bad ticks worse in illiquid stocks?
Yes. Thin stocks and far options get more stray prints and have less trading to correct them, so bad ticks cluster there and the fake fills would have been impossible to trade anyway.
Do daily bars have bad ticks too?
They can. A single wrong print during the day can become a bad high or low on the daily bar, and that false extreme can still trip a stop or a breakout, so daily data needs sanity checks as well.

Sources & references

    Last reviewed 12 July 2026. Educational content only — not investment advice. Markets and rules change; verify current conventions with SEBI, NSE/BSE and your broker.

    ← Previous
    Time Zones & Timestamps
    Educational content only — not investment advice. Examples use illustrative numbers and simplified models. Backtested results are hypothetical and trading derivatives involves substantial risk. See our Risk Disclosure and SEBI Disclaimer.