Research verdicts

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

touch retention

How much of a signal survives when it is measured on one side of the book alone instead of the midpoint. A signal that mostly vanishes was mostly bounce.

bid-ask bounce

A fake price wiggle caused by trades alternating between the buy price and the sell price. The item's value has not moved; only which side traded last has.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

cohort

A group of items fixed in advance, so the grouping cannot be adjusted after the result is seen.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

pooled

One number computed over every item's rows thrown together. A pooled figure can look strong while the per-item figures are weak, which is why both are always reported here.

null

The dull baseline a model has to beat, such as assuming tomorrow looks like today. It is chosen and published before any model runs, so it cannot be picked afterwards to make a result look better.

notional-weighted

Two ways of averaging across items. Equal-weighted counts every item the same. Notional-weighted counts the items with more money moving through them for more.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

touch_only

A deliberately generous rule for deciding an order filled: it counts a fill whenever the market price reached the order's price at all. Real fills are a subset of these, so a result measured this way is an upper bound on what could have happened, not an estimate of what did.

attempted cycle

One full attempt at the strategy: place a buy order, wait, and sell if it fills. Attempts that never filled are counted too, so an average per attempted cycle cannot be flattered by dropping the ones that did not work.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

drift share

The fraction of a result explained by the whole market moving rather than by the strategy doing anything.

DRIFT_SHARE_THRESHOLD

The cutoff, fixed in advance at 0.5, for how much of a result may come from the whole market drifting rather than from the strategy earning it. Above the cutoff the result is not credited.

residual

What is left of a return once the known costs and effects are subtracted. It describes how the quote was priced and is not earned edge, which is why it no longer decides whether a strategy passes.

touch_only

A deliberately generous rule for deciding an order filled: it counts a fill whenever the market price reached the order's price at all. Real fills are a subset of these, so a result measured this way is an upper bound on what could have happened, not an estimate of what did.

bps

Basis points. One basis point is a hundredth of a percent, so 100 bps is 1 percent.

weighted quote

The headline price Hypixel reports, averaged across the orders on one side of the book rather than taken from the best one. It sits wider than the touch, about 1.8 percentage points at the median, so a spread measured against it overstates the real cost of crossing.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

attempted cycle

One full attempt at the strategy: place a buy order, wait, and sell if it fills. Attempts that never filled are counted too, so an average per attempted cycle cannot be flattered by dropping the ones that did not work.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

attempted cycle

One full attempt at the strategy: place a buy order, wait, and sell if it fills. Attempts that never filled are counted too, so an average per attempted cycle cannot be flattered by dropping the ones that did not work.

touch_only

A deliberately generous rule for deciding an order filled: it counts a fill whenever the market price reached the order's price at all. Real fills are a subset of these, so a result measured this way is an upper bound on what could have happened, not an estimate of what did.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

cohort

A group of items fixed in advance, so the grouping cannot be adjusted after the result is seen.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

grid_edge_unidentified

The best setting the search found sat on the outer edge of the range that was searched. That means the edge of the search chose it, not the data. Widen the range and the answer would move, so this is not an answer yet.

vol_7d

A ranking of items by how much they traded over the previous seven days.

attempted cycle

One full attempt at the strategy: place a buy order, wait, and sell if it fills. Attempts that never filled are counted too, so an average per attempted cycle cannot be flattered by dropping the ones that did not work.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

null

The dull baseline a model has to beat, such as assuming tomorrow looks like today. It is chosen and published before any model runs, so it cannot be picked afterwards to make a result look better.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

depth_bps

How far below the going rate a buy order is placed, in basis points. Deeper means a better price if it fills, and a smaller chance it fills at all.

sell_depth_bps

The same setting on the way out: how far above the going rate the sell order is placed once the buy has filled.

ttl_days

How many days an order is left sitting before it is given up on. This is the patience setting.

grid_edge_unidentified

The best setting the search found sat on the outer edge of the range that was searched. That means the edge of the search chose it, not the data. Widen the range and the answer would move, so this is not an answer yet.

canonical digest

A fingerprint of the exact numbers a verdict was written against. If the underlying result is ever regenerated and changes, the fingerprint stops matching and the claim no longer applies.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

touch_only

A deliberately generous rule for deciding an order filled: it counts a fill whenever the market price reached the order's price at all. Real fills are a subset of these, so a result measured this way is an upper bound on what could have happened, not an estimate of what did.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

COVERAGE_GAPPED

A status meaning the recording has holes in it. Anything built on top has to stop at each hole rather than guess across it.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

holdout

A slice of the data set aside and deliberately never looked at, saved for one final test. Looking at it early spends it, and it cannot be put back.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

negative stop

The measurement came back negative and the work stopped there, by a rule agreed before the measurement ran. It was not retried under different settings until it came out differently.

fill-hazard

An attempt to predict how likely a resting order is to fill in the next stretch of time, so that orders unlikely to fill could be skipped.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

REQUIRES_ACQUISITION

A status meaning the data needed does not exist yet. The question cannot be answered until it is recorded.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

negative stop

The measurement came back negative and the work stopped there, by a rule agreed before the measurement ran. It was not retried under different settings until it came out differently.

holdout

A slice of the data set aside and deliberately never looked at, saved for one final test. Looking at it early spends it, and it cannot be put back.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

non-candidate

Ruled out as something to build on. Later stages that depended on it do not get to run on a weaker substitute.

fill-hazard

An attempt to predict how likely a resting order is to fill in the next stretch of time, so that orders unlikely to fill could be skipped.

negative stop

The measurement came back negative and the work stopped there, by a rule agreed before the measurement ran. It was not retried under different settings until it came out differently.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

non-candidate

Ruled out as something to build on. Later stages that depended on it do not get to run on a weaker substitute.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

cohort

A group of items fixed in advance, so the grouping cannot be adjusted after the result is seen.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

withheld

The work was done and the result was not published. That is a different thing from abandoned and a different thing from never attempted.

weighted quote

The headline price Hypixel reports, averaged across the orders on one side of the book rather than taken from the best one. It sits wider than the touch, about 1.8 percentage points at the median, so a spread measured against it overstates the real cost of crossing.

interior optimum

A best setting that sits inside the range searched rather than on its edge. Finding one is what shows the data picked the setting. Not finding one is why the search above is not treated as an answer.

canonical digest

A fingerprint of the exact numbers a verdict was written against. If the underlying result is ever regenerated and changes, the fingerprint stops matching and the claim no longer applies.

This site's own explanation of the words above. The verdict itself is published exactly as the research index records it and is not rewritten here.

the touch

The best price sitting on the order book right now. It is not the same as the headline price the game shows, which is an average across the orders on that side.

price-bound artifact

A fill the simulation counted only because the recorded price range touched the order's price, with nothing else showing a trade actually happened. It is a measurement error, not a fill.

VOLUME_CONFIRMED

A stricter fill rule that also requires traded volume to back up the price move. Stricter did not turn out to mean more accurate: it was measured carrying a higher share of false fills than the generous rule at every depth tested.

touch_only

A deliberately generous rule for deciding an order filled: it counts a fill whenever the market price reached the order's price at all. Real fills are a subset of these, so a result measured this way is an upper bound on what could have happened, not an estimate of what did.