Scarb Glossary
live

Arbitrage, anomalies and dislocations in crypto

Where prices come apart, why, and what that is actually worth. We write the analysis and we run the monitors and scanners behind it — the two are the same work: a tracker that nobody can explain is a screenshot, and an explanation nobody measures is an opinion.

The events: an exchange removes a coin, a coin arrives on an exchange, locked tokens come free, withdrawals close, a stock contract drifts from the stock, one coin trades at two prices, a protocol gets drained.

Plus the part nobody usually publishes: how we check an idea, and the list of ideas that failed the check.

11
venues polled
7
event types
45s
scan cycle
38
pages

Figures in the watch blocks are placeholders until the pages are wired to the live store.

How dislocations happen

7 articles

The mechanism: where the event comes from and what people do with it

An exchange removes a coin
date known in advance
Every holder gets a deadline, and there are no buyers on the other side of it.
A coin arrives on a large exchange
first week is noise
Day one puts the people buying the news in the same book as the people who got the tokens for nothing.
Locked tokens come free
schedule is public
New supply arrives on a known date, and gets sold afterwards, slowly.
An exchange closes deposits or withdrawals
checked continuously
While the door is shut, the price there drifts away from everywhere else — there is nothing to move.
A stock contract against the stock itself
new format, few cases
The stock sleeps on its exchange's schedule; the contract trades around the clock.
One coin, different prices
the question is capacity
There is no common price — only what is resting in one particular book.
A protocol gets drained
no date beforehand
The only event here that you learn about afterwards rather than in advance.

How we verify

10 articles

Ways to tell a real find from a coincidence

Test it on a different coin
The main check for everything: same minute, same side, different coin. If the result doesn't change, the rule had nothing to do with it — the market did the work.
Split the window in half
Cut the stretch in two and count each half on its own — the cheapest way to catch a made-up result.
Never pick the best setting from the past
Try a hundred settings and keep the best one, and you don't have a rule — you have a description of the past.
Check liveness by the data, not by the code
Switched on and actually writing data are two different claims, and only the data can settle the second one.
The share matters more than the average
On scattered data the average barely means anything — what you want is the share of cases that finished up.
Profit per trade is not a business
A rule can win on every single trade and still lose to just buying and holding — because the money has to be split.
Same units before you get excited
Half of the loud gaps between venues turn out to be a difference in units of measurement, not in price.
Measure liquidity in the book, not in daily volume
Daily volume tells you how much trading happened, not how much money the market will take right now. The difference can be several times over.
The exit rule makes the result, not the entry
Before you praise anyone's entries — someone else's or your own — run the same exit rule on random entries. Very often the whole result lives in the exit.
Compare to the market over the same stretch
Comparing to the market only means something over the same stretch of time the strategy ran. Otherwise you can get any number you like.

Tested — does not work

21 articles

Ideas that sounded convincing and did not hold up

A long shelf, then a break upward
The idea: price sits in a well-worn range for months, and leaving it means a run to the next such range. When we tested it, breaking out of the shelf paid exactly as much as a random day on the same coin.
Big orders in the book as a reference point
The idea: a big order in the book is a level price will bounce off, or a level it will aim for. When we tested it, price reached such a wall no more often than it reached a randomly picked level at the same distance.
Buying into a vertical spike
The idea: the coin has gone vertical, so a move has started and you should get on board while it runs. When we tested it, buying the spike was not a zero but a loss: far fewer trades finished up than with a random entry.
Shorting the spike, and the hunt for a better exit
The idea: if buying a vertical spike loses money, selling into it should make money, and all that is left is finding the right exit. We swept the whole space of exits and hit a ceiling that is a zero even in the best case imaginable.
A second signal in a row as confirmation of the first
The idea: do not trade the first signal, wait for a second one on the same coin — it confirms the move is real. When we tested it, the second signal had a lower share of trades that finished up than the first.
Open interest jumping with aggressive flow as a signal
The idea: a sharp jump in open interest together with a price move shows money entering a small coin, so you should follow it in. Tested across several different months, both sides of the trade came out near zero after costs.
Alerts about big scheduled buys on-chain
The idea: you can see somebody place a large buy order split into equal slices, and the promised price impact is enormous. When we tested it, the real move over the order's lifetime was smaller than claimed by orders of magnitude, and there is nothing tradable in the alerts.
A deep drawdown plus a failed spike up
A recognisable bearish chart shape: the coin fell far from its yearly high, then bounced sharply and gave the bounce back. The idea is that this picture tells you something about what comes next. The test showed it tells you nothing.
A dense zone above price as a ceiling
The idea: "do not buy above this price, there is an overhang of other people's orders and other people's trades up there." Tested across a basket of coins, dense zones came out better than thin ones, not worse.
Farming rewards by locking a coin on an exchange
"Lock the exchange's coin, get a new token for free" used to pay noticeably well. By now the income has almost vanished.
A delisting on several exchanges at once
When several exchanges pull the same coin from trading at once, it looks like the strongest sell signal there is. But sometimes it is not a delisting at all, it is a rename, and then there is no signal.
The pressure a sliced order puts on the book
A large order that a bot chops into pieces and feeds into the market really does push the book around. You just cannot get paid for that push: the move does not come back, and what is left to take is smaller than the fees.
Riding along with a big sliced order
The idea: you can see a whale feeding a huge order into the market in pieces, so get in beside him and ride. It did not survive a long test: what is left is smaller than the fees, and all of the profit turned out to be one single month.
When a venue's fair price drifts away from where trades print
The idea: the last trade ran away from the exchange's fair price, and the exchange's own machinery will drag it back. In a calm market it gives you nothing: almost all of those gaps are either untradeable or close in the opposite direction from the one you bet on.
Breakout from a tight range after a long consolidation
The classic move — price sat in a tight range for a long time, left it, so you go with the exit — turned out not to be a property of the coin at all. It is a way of guessing the moment when the whole market moves.
Buying the exit from a range to the upside
Half of the range-breakout rule — buying the exit upward — turned out to be indistinguishable from simply buying bitcoin at the same moment. The whole difference was sitting in the sells.
Rejection off a level and a break of a falling trendline on small caps
The two most recognizable chart moves — selling a touch of a level that does not break, and buying the break out of a falling trendline — gave nothing on small caps, neither against a random hour nor against the market in that same hour.
Selling bitcoin's three-day high when it falls back under
The idea: price poked through the high of the last three days and came back, so we sell. We tested four versions of it, and not one of them beat a random entry in the same window.
Exchange news and token unlocks as a reason to close a short early
The idea: hold the short, but get out early if the coin has exchange news coming or tokens coming out of vesting. It did not hold up: those events do not explain the violent moves against a short, and in one case the news is a reason to hold, not to leave.
The launch of a perpetual contract on a stock as an event to trade
The idea: carry the working move of shorting a fresh listing over from coins to perpetual contracts on stocks. It did not hold up: a contract like that has nowhere to detach to, because it is tied to the real price of the stock.
Tokenized equities in a general scan of coins
Scanning the whole universe of contracts and sorting out the hits later does not work: tokenized equities flood any such scan, and as a trade they give nothing.

Glossary

27 terms
Glossary
Order book, liquidity, funding, float, oracle, mark price — in plain English.