Slipstream & Signal Quality
Slipstream is a spotlight, not a crystal ball. It watches where the biggest brokers are quietly concentrating their money and highlights the stocks worth a closer look, calibrated for each stock on its own. Use it to find candidates, then confirm them with the other signals. Below we explain exactly how it works, and we are honest about what it can and cannot do.
What it shows
Every trading day, Slipstream reads three things about a stock and blends them into one line. The first is order blocks: price zones where a large move once began. Think of them as the spots on the chart where big money stepped in before, the places price tends to return to and test. The second is candle conviction: how decisively price moved. A large, clean candle on heavy volume shows real intent; a small, indecisive one does not. The third is top-broker flow: the net buying of the biggest broker desks, the clearest footprint of institutional activity.
When all three line up, Slipstream "fires" for that day. And because every stock trades differently, we tune it to each stock's own history, like a tailored suit instead of one-size-fits-all. That tailoring is about fitting the signal to the stock, not about ranking one stock above another: what a fire is worth depends on the structure and market context around it, which the next sections cover.
How it is built
three reads, blended into one signal
Reading the calibration state
Signal Calibration tells you that this stock's Slipstream was tuned on its own history and then tested on periods the model had not seen. It describes the calibration. It is not a score, and nothing in it forecasts a return.
For MDKA it reads active: calibrated, liquid enough to act on, and switching on regularly. All three are things you could recount from the data we store. It is NOT a forecast, and it does not mean this stock's signal is more reliable than another's: the differences between tickers are too small to separate from chance.
There used to be a grade here, running from elite down to skip. We withdrew it. We measured whether one stock's signal really does better out-of-sample than another's and could not separate the differences from chance, so any grade we published would have been noise wearing a badge. What replaced it are five plain facts, none better than another: Active means the signal is calibrated, the stock is liquid enough to act on, and the signal switches on regularly. Quiet is the same, except it rarely switches on. Provisional means the stock listed recently, so the settings come from a shorter history. Illiquid means the stock trades too thinly for the signal to be worth acting on. Not enough data means too few tested trades to say anything at all.
Signal Calibration
Parameters are fitted on MDKA's own history, then tested on periods the model has never seen. This describes the calibration — it is not a grade and not a forecast of how the next fire turns out.
Calibration states
descriptions, not a ranking
- Activeliquid, tested, and fires regularly
- Quietliquid and tested, but rarely fires
- Provisionallisted recently — short history
- Illiquidtoo thin to trade the signal
- Not enough datatoo few tested trades to say
How reliable is it
Here is the part most tools would bury. When we tested Slipstream rigorously, we caught our own optimism. One of its components looked like a huge edge, until we found a look-ahead bias hiding in it. A look-ahead bias is like grading a student on a test after they have already seen the answer key: they look brilliant, but the result is fake, because they "knew" something they could not have known at the time. Once we removed it, that component's edge collapsed from +717 basis points to +11 (100 basis points is 1%).
What is left is real but modest, and barely beats plain price momentum. Slipstream is also long-biased: it only ever looks for buying, never for selling short. So in a falling market it keeps pointing up into the decline, which is exactly what happened in the 2026 downturn. That is why we call it a spotlight for where to look, not an engine that predicts returns.
By year, honestly
helped in 2025, inverted in 2026
In short
- 717 → 11 bps once the leak was removed
- Long-biased: a spotlight, not a predictor
- Confirm with structure and the market regime
On the price chart
On a stock's page, Slipstream shows up in two places that sit together. On the price pane itself there is a flow line: a rolling sum of the last fifteen days of top-broker flow. Unlike a running tally that only ever climbs, it rises when recent smart-money buying is building and falls when that flow fades, so it turns with the money instead of drifting up through every decline. When the signal is live the line takes on a firing tint, and a small dip-signal badge marks the days smart-money flow jumped while price was falling, a historically rewarding place to look. Read its slope and direction, not its exact number, which is scaled per stock and not comparable between names.
Below the price sits the momentum pane, and this is the real trigger. It plots near-term buying momentum in standard-deviation units against a dashed Fire ≥ line, the stock's own calibrated threshold, with a dot on each crossing. When momentum crosses above that line the signal fires and a marker pins the day; a small badge strip names the state in words, whether it is firing, accumulating but fading, diverging from price, or just crossing the line. The verdict card carries the same read in a Slipstream block: a Firing or Dormant pill, how far momentum sits above or below the fire line, which way it is sloping, and whether flow and price agree. That crossing is the exact event the grade above is measured on.
MDKA price chart
20 Slipstream fires in view
MDKA is graded active tier, calibrated on its own history (model v417). The green markers show where Slipstream fired during this accumulation run: a trigger to confirm against structure, not a promise.
The crash-regime guard
One honest failure mode earns its own guard. Slipstream reads the biggest brokers as buyers, but in a fast crash that reading can flip on you: when a name is in free fall, heavy broker buying is often just crash-buying, forced or reflexive or bargain-grabbing, not the quiet accumulation the frozen smart-money labels keep scoring it as. Refreshing those labels does not fix it. So we add a blunt rule. When a stock is down more than 30% over the trailing 60 trading sessions, we treat it as a crash regime and stop trusting the accumulation read. The firing tint drops off the flow line, the Firing and Accumulating badges disappear, and the verdict card shows a Crash regime pill in place of Firing. The Slipstream line itself is untouched, only the green "this is accumulation" cue on top of it.
It is deliberately conservative and covers only the steepest declines. Names younger than 60 sessions use a stricter fallback rule: only a collapse of more than half from their post-listing peak trips the guard, because young listings often fade 30 to 40% from their peak without it meaning much. The divergence and dip-signal reads are left alone, because those already weigh flow against a falling price. Think of it as a circuit breaker on one specific way a crashing tape can masquerade as accumulation.
How to use it
Start in the screener with the Firing preset. The signal state is a filter, not a sort: use it to drop names you cannot trade anyway (Illiquid, Not enough data), then lean on the next two steps. It does not tell you which fire is better, because we could not measure that.
Open the chart and check the shape. A fire is strongest when the flow line is already rising and momentum has pushed well above its fire line: pressure building and the trigger confirming together. A lone fire while flow is falling is weak.
Confirm it with structure and the market. A fire inside a Wyckoff accumulation phase, while the broad market is not in a downturn, is far stronger than a fire in mid-air. This is the most important step, because Slipstream on its own is weak.
Decide your exit before you enter. Set your stop on the structure and let momentum turning red be one cue to trim. Size by your own risk limits — the signal state says nothing about how much to risk.
Weak on its own, and long-biased. On a wild name or a falling market it can keep firing. Pair it with structure and the Market-Risk gauge so the market's weather is never a surprise.
Selective by design. On a typical day only a few dozen of the roughly 795 calibrated stocks are firing. No fire is not a sell signal, it just means there is no setup on that stock today.
Uncalibrated names have no state yet. A stock gets one only once it has enough history to fit and test on. Until then, lean on the other reads on the page.