Open the Prediction Alpha markets table and look along a row. The left half is dense with numbers. The right half, from Signal onwards, is a run of small grey dots. The natural reading, and the wrong one, is that the platform looked at this contract and found nothing worth flagging.
It did not look. That is a completely different fact, and confusing the two is how people end up trading the wrong thousand markets. A blank is a statement about coverage, and coverage is a property of the queue rather than of the contract.
The row is two halves with different origins
Read the columns in order and the split is obvious once you see it. Market, Yes, No, 24h Vol, Tot Vol, Liquidity, End Date. Then Signal, Misp%, Kelly%, Whales, Traders. Then an Analyze button and a Trade button.
Everything before End Date is exchange data. It exists because the market exists and somebody traded in it, and it arrives whether or not anyone has an opinion. Everything after End Date is derived: a signal is a judgement, a mispricing percent is a comparison against an estimate, a Kelly percent is a sizing output of that estimate, and the participant counts come from a wallet identification effort that has to have run.
Derived numbers have a production cost. Exchange numbers do not. That is the entire reason one half of the row is always full and the other half usually is not, and it is worth understanding because it tells you which absences to worry about.
The Analyzed chip is the coverage number, and it is small
The filter bar in that capture carries a set of chips, and one of them reads Analyzed with a count of 10. Above the table the header tiles read 56.2K markets indexed across six venues, with 1.0K displayed after filters. Two other chips, Cross Match at an 80 percent threshold and Arb Only, both carry a count of 79.
Put those together. Ten analysed markets against a thousand displayed is one percent. Against the indexed universe it rounds to nothing. And each row carries its own Analyze button, which tells you the analysis is something you request rather than something that has already been run everywhere and come back negative.
So the correct translation of a blank Signal cell is: nobody has asked this question yet. Not: the answer came back boring. The first is an invitation, the second is a verdict, and the screen is only ever giving you the first.
The same logic applies to the other two chips, and it is worth saying because they look like market facts rather than coverage facts. Seventy nine cross matched markets at an 80 percent threshold does not mean seventy nine questions exist on two venues. It means seventy nine pairs cleared a similarity score inside the thousand rows currently displayed. Change the filters above the table and both numbers move, which is the clearest possible demonstration that these counts describe the screen rather than the world.
The gaps you can trade straight through
A blank derived half on a row whose exchange half is rich is close to harmless. Look at the top row of the screenshot: the Clarity Act market, quoted 18.5 against 81.5, with 1.9M of 24 hour volume, 10.0M of total volume, 497.3K of resting liquidity and an end date of 01/01/27.
Nothing you need is missing there. The price is a genuine crowd number, produced by ten million dollars of accumulated trading rather than by a handful of prints. The book will take a retail sized order in and back out. The resolution date is a long way off, which is a capital commitment question but not an information question. If you have a view on that legislation, you can form it, size it and act on it without a single derived column, exactly as you would if the platform did not compute them at all.
That is the honest test for whether a gap is safe: would you still take this trade if the derived columns did not exist as a feature. If yes, the blank costs you convenience and nothing else. Press Analyze if you want a second read, but the trade does not depend on it.
The gaps that should stop you
Now the second row. Gunnar Henderson to lead the MLB in runs, quoted 0.9 against 99.1, with 796.4K of 24 hour volume against 796.8K of total volume, and 686 dollars of resting liquidity.
Here the blank compounds a problem rather than sitting beside one. There is no derived read, and there is no usable crowd read either, because essentially every dollar that has ever traded in this market traded in the last day. A price of 0.9 percent produced by a book of 686 dollars is not a probability estimate in any meaningful sense, it is wherever the last few orders happened to leave the quote. And your exit is that same 686 dollars, so a 250 dollar position is more than a third of everything resting.
Two blanks are disqualifying rather than merely inconvenient. When the derived half is empty and the exchange half is thin, you have no read at all from any source, and no amount of clicking Analyze fixes a book that small. And when the Signal cell was going to be your reason for the trade, the blank has just told you that you did not have a reason. That second one sounds obvious written down and it is remarkably common in practice: people arrive at a market through a signal filter, find the signal missing, and take the position anyway on the momentum of having already decided.
The screening habit that follows from all this
Stop hunting for populated rows. Filtering to Analyzed gives you ten markets selected by the fact that somebody looked at them, which is not correlated with them being good and is mildly correlated with them being crowded. Letting a coverage chip define your universe is the single easiest way to end up trading a queue.
Do it the other way round. Sort by 24 hour volume, apply your own liquidity floor, read the exchange half on the twenty or thirty rows that survive, and pick the two or three questions you actually have something to say about. Then press Analyze on those. You are using the derived columns as a check on a view you already formed, which is the only role they can safely play when they cover one percent of what is displayed.
And write the coverage number down somewhere you will see it. Ten of a thousand. It is a useful antidote to the feeling, which every well built screen creates, that the empty cells are telling you something.