summary: why fast markets — especially the friday close — cause outsized slippage on automated algos, and the concrete settings changes that keep a volatile session from turning into a max-loss day.
your algo doesn't know it's friday. it fires the same setup, the same size, into a market that can be moving twice as fast with half the liquidity. that mismatch is where most "friday disaster" stories come from — not a broken algo, just an algo running an average-day config on a not-average day.
here's what's actually happening, and what to change so a fast session doesn't cost you more than you planned.
why fast markets hurt an algo more than you
slippage is the gap between the price your alert fired at and the price your broker actually filled. it's almost always negative on exits, and it gets worse in exactly two conditions: speed and thin liquidity. fast fridays, the cash open, and news windows hit both at once.
two things stack up along the way:
latency — there's a chain from your TradingView alert to edgeful to your broker (around 300ms on average). in a fast tape, price can travel several ticks in that window.
liquidity — a market order fills across whatever's resting in the book. when the book is thin, one order sweeps multiple price levels and your fill lands well past your level.
and your stop-loss is always a market order — it fires the moment price hits your SL level, at whatever the market is right then. that's the one you can't limit-order your way out of, so it's where fast-market slippage does the most damage. the full mechanics are in execution, slippage, and order types in algo automation.
the friday problem specifically
friday adds two things a normal session doesn't:
the weekly close — liquidity thins out into the afternoon as desks square up for the week, so fills late in the friday session are often the worst of the week.
a position you don't want to carry — edgeful algos flatten at the end of your trading-hours window, but if that window runs into the fastest, thinnest part of the friday close, your exit fills there.
none of that means the edge is gone. it means the cost of trading the tail end of friday is higher than the cost of trading a tuesday morning — and you can adjust for it.
what to change on volatile days
these are the levers you already have in your algo settings. you don't need a new strategy — you need a version tuned for the conditions.
trim your trading hours
the single highest-leverage change. shorten your trading hours window so the algo isn't taking new entries into the fastest part of the session — and so its end-of-day flatten lands before liquidity falls off, not during it. how trading hours and the flatten work: algo parameters: take-profit, stop-loss, trading hours, and risk controls.
size down into the windows you know are fast
fewer contracts means every tick of slippage costs less, and a bad fill can't blow past your risk. if friday is consistently your worst day in your own results, that's your data telling you to run it smaller — or sit it out.
reconsider market vs limit on fast entries
market orders guarantee you're in but accept the slippage. limit orders control your price but may not fill at all when price rips through your level — which on a fast friday is exactly when they miss. there's no universally right answer; test both on SIM and pick the one whose failure mode you can live with on your setup.
treat max loss as a circuit breaker
your max loss setting is what stops one fast session from becoming a blow-up. set it to a number you're willing to lose on a bad day and let it do its job — it's the backstop for exactly the friday scenario. on a funded or drawdown-limited account, consider also placing a manual broker-side stop on the position, since the algo's stop isn't a resting order parked at your broker.
this is variance, not a broken algo
one ugly friday inside the drawdown your backtest already showed is not a reason to tear up your settings. changing the algo in the middle of a normal-but-painful stretch is how a losing streak becomes a permanent decision. the line between normal variance and an actual problem — and when to leave it alone — is in what to expect from algos: drawdowns, losing streaks, and when to change settings.
