A statistic is a summary of a period. It carries the conditions of that period inside it. When the conditions change, the number is still an accurate summary of the past and a poor description of the present.
What a regime is
A regime is a stretch of time in which a market behaves in a broadly consistent way. Volatility is low or high. Rates are rising or falling. Dealers are positioned so that their hedging dampens moves, or so that it adds to them. Within a regime, averages and ranges are fairly stable. Across regimes they are not. Volatility regimes is the clearest example.
Why statistics assume the regime
Any average quietly assumes that the future is drawn from the same conditions as the sample. Statisticians call that stationarity. Markets are not stationary for long. The typical daily range of an index in a calm year tells you little about its range during a crisis, and a rule tuned on one will meet the other eventually.
This is why a tested rule can stop working without anything being wrong with the test. The data was real, the arithmetic was right, and the conditions that produced the pattern ended. More data from the old regime would not have helped.
Changes that are structural
Some changes are cyclical and come back. Others are structural and do not. The growth of same-day options changed how much of the options market expires each afternoon. Index perps that trade through the weekend mean a live price now exists on Saturday where none did before. A statistic measured before a structural change describes a market that no longer exists in that form.
New markets raise the same problem from the other side. A perp that listed this year has lived through one set of conditions only. Its whole history may be a single regime, a point made in a market with no history cannot be backtested.
How to tell the regime moved
There is no bell. The usual evidence is that several measured things shift together and stay shifted: realized ranges, the level of implied volatility, correlations between assets, the sign of dealer gamma. One odd day is noise. A month in which the old averages keep missing in the same direction is a different matter.
A practical habit is to keep two windows side by side. A long window gives the stable picture. A short window shows what is happening now. When they disagree widely, the question of which one describes today is worth asking out loud. What a percentile window means covers how the choice of window changes the reading.
Conditioning instead of averaging
One response is to stop pooling everything. A base rate measured only on days that resemble today is narrower and usually more relevant than one measured on all days, as conditional base rates explains. The cost is fewer cases, so the number is less stable. There is no free answer. A broad sample mixes regimes and a narrow one is small.
Where NoVo fits
NoVo reports the present regime as a measurement. Trader shows dealer positioning on SPY, QQQ and IWM, including net GEX and the gamma flip, so a reader can see which side of the flip price is on today. It reports what is measured now and leaves the forecast out.