Curriculum·R409 Grid, DCA, and Rule-Based Automation·about 33 min
Regime dependence, which is the whole game
By the end of this lesson you can
- →State why a grid has no expectancy independent of regime
- →Show that a backtest on one regime is uninformative about the others
- →Compute the same configuration's result across three regimes and read the spread
- →Name what would have to be true for the configuration to be worth running
Senior · enrolled learners
This lesson opens with The quant quake, 6 to 9 August 2007, read as a regime change.
- What happened
- Over four days in August 2007, quantitatively managed equity market-neutral and statistical arbitrage funds recorded losses without precedent in their own histories, in a week when the broad equity market moved very little. Khandani and Lo attributed the sequence to the rapid unwind of one or more large quantitative portfolios, probably forced elsewhere, which pushed prices against every other portfolio holding similar positions and triggered their own deleveraging rules. Goldman Sachs' Global Equity Opportunities fund lost more than 30 percent that week and received a capital injection. Most of the strategies rebounded sharply on 10 August. The models had not stopped working and the signals had not decayed. The conditions the models required had stopped holding for four days.
- The decision point
- Every one of those strategies had been validated on history and every validation was correct about the history it used. What none of them contained was the state in which the relationship they depended on temporarily did not hold. A grid depends on one such relationship explicitly, being that the price crosses its levels repeatedly and remains inside its range. That is not a parameter the configuration sets. It is a description of the market the configuration requires, and it is the only thing that determines whether the rule earns anything at all.
What you will be able to answer
- →Why has a grid no independent expectancy?
- →What does a one-regime backtest establish?
- →What is the spread across regimes?
- →What would have to be true to run it?
Orientation and Year One are open: anyone can read them without an account. From Year Two onward the lessons are for enrolled learners, because progress through the later years only means anything if it is tracked against a record.
It is free. We do not sell the list and there is nothing to buy at the end of it.
Sources and review
- https://web.mit.edu/Alo/www/Papers/august07.pdf
- https://www.nber.org/system/files/working_papers/w14465/w14465.pdf
- https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1015987
- https://global-q.org/uploads/1/2/2/6/122679606/houxuezhang2020rfs.pdf
Confidence medium·Volatility low·Reviewed 2026-08-07·Owner unassigned
Contested
R405 uses the quant quake at course level for crowding, being that independently validated strategies converged on the same positions. This lesson uses the same episode for regime dependence, being that the conditions a validated rule requires can stop holding without the rule changing. The two arguments are related and distinct, and the split is declared. Keep it.
The three-regime comparison carries forward R409-03's stated price paths and R409-01's configuration so the numbers remain checkable across the course. Per R409-L2 the lab requires the learner to run their own configuration against real ranging, rising and falling periods with fees and funding, and to pass by producing a regime in which it loses badly and explaining why.
R405-03 owns the seven sins including regime selection. R405-04 owns sample size. This lesson owns the specific case where the strategy's premise is a market state. Keep the splits.