Why you can't copy someone else's trading bot
Survivorship bias, edge decay, and the gap between a backtest and a live account.
The concepts behind the indicators and reports, explained plainly and without mystique. If a term shows up in the research, it should be defined somewhere here.
Liquidity, in the trading sense, is simply resting orders — stops and limit orders sitting at prices where people have decided to act. They cluster in obvious places: just beyond yesterday's high, under an evident swing low, around round numbers.
That clustering matters because large participants need volume to fill a position. A pool of resting orders is where that volume is. This is the mechanism behind the pattern retail traders describe as being "stop-hunted" — usually not deliberate targeting of individuals, just size going where the fills are.
Practical consequence: an obvious level is not automatically support. It's a place where orders are concentrated, which makes it a candidate for both a bounce and a break. The 503-session study puts numbers on that — most levels hold roughly half the time on first touch.
A sweep is when price trades through a liquidity pool — taking out the high or low — and then fails to continue. A reclaim is when it closes back inside the prior range shortly afterwards.
The combination is more informative than either alone. A break with follow-through suggests genuine directional interest. A break that immediately reverses suggests the move existed to access orders rather than to establish a trend — the buyers who chased the breakout are now offside, and their exits push price the other way.
This is what the COT (change of trend) and COC (change of character) patterns encode: sweep, then reclaim within a defined window. See it implemented →
Every trade has a buyer and a seller, so "buying volume" needs a precise definition. Delta measures which side was aggressive — whether a trade executed at the ask (an aggressive buy) or at the bid (an aggressive sell). Cumulative Volume Delta sums that over time.
Its use is comparative. When price makes a new high but CVD does not, the move is being driven by something other than fresh aggressive buying — often thin liquidity rather than genuine demand. That divergence is a warning, not a signal: divergences can persist far longer than a position can be held. The indicator →
Market profile organises a session by how much time price spent at each level rather than by time alone. Levels where a lot of trade occurred represent agreement on value. Levels that were passed through quickly represent rejection.
A single print is a price touched in only one time bracket — a fast, one-sided move that left a gap in acceptance. A naked point of control is a session's highest-volume price that has not been revisited since. Both are treated as magnets on the reasoning that unfinished auctions tend to be revisited, though "tend to" is doing real work in that sentence.
The ISM PMI is a diffusion index. It doesn't measure output — it surveys purchasing managers and records the breadth of responses. Above 50 means more respondents report expansion than contraction; below 50, the reverse.
Two implications that get missed. First, the index says nothing about magnitude: 55 means expansion is widespread, not that it's large. Second, direction matters more than level — a fall from 58 to 54 is deceleration even though 54 is still expansionary.
Its value is timeliness. Survey data arrives well before hard data, which is why it anchors the monthly playbook. The PMI playbook →
Building permits sit at the very front of the construction pipeline. A permit is a decision to commit capital, recorded before any material is bought or anyone is hired. Everything downstream — starts, completions, employment, materials demand — follows it with a lag.
Housing also has unusually broad linkages: lumber, building materials, appliances, furniture, mortgage lending, insurance, and a large share of household net worth. Weakness propagates rather than staying contained. That combination — early in the chain, widely connected — is what makes permits worth watching out of proportion to the sector's direct share of GDP. The full argument →
A regime score aggregates several independent signals — growth, labour, rates, credit, breadth, volatility — into a single reading. Each contributes +1, 0 or −1 according to rules set in advance.
The rules-in-advance part is the entire point. Discretion tends to arrive precisely when it's least reliable: after a sharp move, when the narrative is loudest. A pre-set trigger such as "payrolls below 75k downgrades labour" can't be argued with after the fact — it either fired or it didn't.
The trade-off is real: mechanical rules are slower to adapt and will occasionally be wrong in ways a thoughtful human wouldn't be. The judgement is that consistency is worth more than responsiveness. This month's score →
Survivorship bias, edge decay, and the gap between a backtest and a live account.
Sample sizes, stated limitations, and why studies that found no edge still get published.
For educational purposes only. Nothing on this page is investment advice.
One email when the report publishes. No trade signals, no affiliate spam.