Alpha Survey · in search of alpha
Menu

Glossary

The words of trading, explained in prose and with a picture, one at a time. Each term enters the glossary the week a log entry first needs it, so the list grows with the logbook rather than ahead of it.

43 results

The rule, written firstif X then buy · if Y then sellReplayed over the past2010TodayA track record that never existed: evidence only if the rule and the bar came first
Backtest
A simulation of how a trading rule would have done on past prices.
first in H-042
1 bpSpread, 5 bpRound trip, 30 bp1% = 100 bpAn edge of 7 bp per trade against a cost of 30 bp is a losing rule.
Basis point
One hundredth of a percent.
first in H-042
The coin everybody is chasingShort here, at the highest fundingPrice move against the shortSmall fee collected every eight hoursEaten many times over by the move
Bet against momentum
Being short the assets that have been rising the most.
first in H-016
Last month's return, rankedWinners: boughtLeft aloneLosers: sold shortBuy the top onlyRises with the marketTop minus bottomThe market cancels outWhat is left is the picksThe short hand pays rent and its losses have no ceiling.
Betting against losers
Shorting the stocks that fell the most: the second hand of momentum.
first in H-044
Yearly rent to borrow the shares0.4%Most large stocks8%Crowded shorts40%Hard to borrowThe same short, before and after the feeGross (what the pitch shows)Net of the fee, every dayTime held →
Borrow fee
The rent you pay to borrow a share so you can sell it short.
first in H-059
Buy the dip
Buying a stock right after it drops, betting it will bounce.
first in H-042
1 version tried2 versions tried5 versions tried20 versions triedThe bar a result must clearEvery attempt in the family counts, including last month's.
Correction for looking twice
Raising the bar when more than one version of an idea was tested.
first in H-042
Fourteen short sellers, one stockThe stockHigh borrow feeFew shares left to lendOne narrow door out:If the price rises, they all buy back at once
Crowded shorts
Stocks that many people are already betting against.
first in H-059
The measure orders returns12345678910Top decile (highest value)Bottom decileThe measure orders nothing12345678910Top decile (highest value)Bottom decileRank every stock by one measure, cut the list into ten equal slices, and look at each slice's average return.
Decile
One of ten equal groups after ranking stocks by something.
first in H-059
Still listed: stays in the dataGoes to zero, delisted…And erased from most databasesThe average you are shown counts only the green line.
Delisted
A company removed from the exchange, and usually from the databases too.
first in H-059
Every eight hours, the crowded side pays the other00:00Day 2Day 40.01% per 8 hours ≈ 11% per yearFattest right before the crowd gets squeezedLongs pay when the contract sits above the coin; shorts pay when it sits below
Funding rate
The fee one side of a perpetual pays the other every eight hours.
first in H-016
Average gain per trade, before any cost12 bpSignal A22 bpSignal B50 bpSignal CToll per round trip: 35 bpAbove the toll: only green is keptBelow it: every trade loses moneyThe pitch shows the whole bar; you keep only the green.
Gross edge
What a signal earns per trade before paying any cost of trading.
first in H-042
Low turnover0.5% of shares traded per dayHigh turnover5% of shares traded per dayA measure of attention, not of size. Computed two ways, it can make a published effect appear or vanish.
High turnover
Stocks that trade a large share of their existing shares each day.
first in H-044
Company AMember throughoutCompany BWent bustCompany CJoined after a runCompany DMember throughoutCompany EBought outCompany FJoined recently2005201020152020TodayToday's list, used for the past, drops B and EAnd counts C and F years before they ever joined
Index membership
Which companies were actually inside an index on each past date.
first in H-042
Large caps: liquid, cheap to trade, crowded with professionalsSmall names: where published effects usually live, and where a real account struggles to trade
Large caps
The biggest companies on the market by value.
first in H-042
Liquid nameSellersBuyersSpread 1 bp: deep queues on both sidesAn order of 1,000 shares barely moves itThin nameSellersBuyersWide gapSpread 60 bp: few orders, far apartThe same order walks the price up 2%
Liquid names
Stocks that trade heavily every day, so an order barely moves the price.
first in H-044
A market that rose 10%The market, +10%+12%basket you ownLong+10%basket you shortShortWhat is left is the bet+2%long minus short+12 − 10≈0If the selectionDid nothing
Long-short portfolio
Owning one basket of stocks while betting against another at the same time.
first in H-059
A market that rose 10%The market, +10%+12%basket you ownLong+10%basket you shortShortWhat is left is the bet+2%long minus short+12 − 10≈0If the selectionDid nothing
Market neutral
A setup built so the market's direction cancels out of the result.
first in H-016
Momentum
Betting that what has been rising keeps rising, and what has been falling keeps falling.
first in H-044
Chance that at least one version passes by luckOne honest look: 5%123%540%1064%2087%40About two in three by twenty triesVersions of an idea with no edge, tried one after another
Multiple testing
Try enough versions of an idea and one will look good by luck.
first in H-044
Gross return +40 bpCommission -5 bpSpread -10 bpSlippage -8 bpBorrow fee -12 bpNet: +5 bp. The chart you are shown stops at the first bar.
Net return
What is left after paying every cost of trading.
first in H-042
In sample: where the rule was builtOut of sample: sealed, opened onceTuned, chosen, discarded against these yearsThe only number that decides
Out of sample
The stretch of history the strategy never saw while it was designed.
first in H-016
The coin (solid) and its perpetual (dashed)Contract above the coin: buyers pay fundingNormal future: expires herePerpetual: no end date, funding keeps it close
Perpetual futures
Contracts that track a coin's price and never expire.
first in H-016
Post-earnings drift
The tendency of a stock to keep moving in the direction of its earnings surprise for weeks.
first in H-042
Before the run · datedThe rule, exactlyThe data and the costsPass if t ≥ 2 and p < 0.025Fail otherwise. No changes after.After the runt = 0.69 → failNo "let's try t ≥ 0.5 instead"No new side picked from the dataThe freedom to adjust afterwards is what manufactures strategies fitted to the past.
Pre-registration
Writing the rule and the pass or fail bar before looking at the result.
first in H-042
ZeroPaper publishedIn the paper's sampleAfter publicationThe years the paper measuredLarge companies today: close to zeroIllustrative: many anomalies weaken once they are traded
Published anomaly
A market pattern written up in a research paper, and often traded away since.
first in H-042
Quarterly results
The report a company publishes every three months on how its business did.
first in H-042
Large fund6 bp per round tripSmall account32 bp per round tripFixed pieces of the cost do not shrink with the order; the edge has to be larger than the toll.Every logbook number is net of these, and reported again with costs doubled.
Real costs
Every fee, spread and slippage a trade actually pays, as a small account pays it.
first in H-042
month 1Heldmonth 2Heldmonth 3Heldmonth 4Heldmonth 5On each date: rank the universe again, sell what no longer qualifies, buy what now does.Every rebuild costs commissions, spreads and, on the short side, borrow fees.
Rebalance
Rebuilding the portfolio on a fixed schedule, here once a month.
first in H-059
You borrow and sellPosition open, paying the borrow feeLender recalls the sharesYou must buy back today, at this price…Where you had planned to close
Recall
When the lender demands the borrowed shares back at any moment.
first in H-059
Relative strength ranking
Sorting stocks by how much they rose compared with the others.
first in H-044
7030"Overbought" above 70, "oversold" below 30Arithmetic on past prices: it holds no information the prices did not already hold. It earns a place only by passing a test.
RSI indicator
A gauge that labels a stock oversold or overbought from its recent ups and downs.
first in H-042
Same return, two very different ridesBoth end hereSmooth: high Sharpe, the return was earnedWild: low Sharpe, the return was survivedSharpe = return above cash ÷ how much it swung
Sharpe ratio
Return earned per unit of risk taken, the usual yardstick for a strategy.
first in H-016
16 Of 80 shares sold short: short interest 20%Every red square must be bought back one day: the fuel of a squeezeDays to cover = short interest ÷ shares traded on a normal day
Short interest percentage
The share of a company's stock that is currently sold short.
first in H-059
You sell borrowed shares at 100…100Sell (borrowed)buy back at 80: +20The gain stops at zero…and the price rises100Sell (borrowed)buy back at 160: −60The loss has no ceiling
Short selling
Betting that a stock will fall instead of rise.
first in H-059
Short side
The half of a long-short portfolio that bets on prices falling.
first in H-042
2,000 shuffled versions of the same strategyReal result: p 0.37, inside the crowdWhere a real edge would sitEverything is kept, market, calendar, costs, except the selection.
Shuffle test
Scrambling the signals to see whether the real ones beat random ones.
first in H-044
Short seller closesAnother one closesAnd anotherHeavily shorted stockThe chart you are shownPrice risesShorts loseForced to buy
Squeeze
When people betting a stock falls must buy it back, pushing the price up.
first in H-059
The universe as it was40 companies, 9 of them gone by todayWhat a download shows31 survivors, none of which went to zeroMeasured on our data: the survivors' Sharpe is inflated by more than half.The fix is to rebuild the list date by date, dead included.
Survivors only
Testing on the companies that are still around today, and missing the ones that died.
first in H-042
Month-to-month results, tightZeroSame averagesmall wobble → t above 2: signalMonth-to-month results, looseZeroSame averagelarge wobble → t below 2: cannot be told from noise
t statistic
A number saying how far a result stands from pure noise.
first in H-059
Top tercilebest reactionsMiddleBottom tercileworst reactionsCoarser than deciles: three groups stay populated where ten would be almost empty.
Terciles
Three equal groups after ranking by something.
first in H-042
2,000 shuffled versions of the same strategyReal result: p 0.37, inside the crowdWhere a real edge would sitEverything is kept, market, calendar, costs, except the selection.
The null
The plain benchmark a strategy has to beat before anyone can say it works.
first in H-044
Thousands of names: the effect has roomThe stocks that carry it are inside the fenceA few hundred large names: one carrier leftSame rule, different answerSameRule
Universe
The list of stocks a test is allowed to choose from.
first in H-059