Adverse Selection, Last Look, and Toxic Flow: Tricks of a Defensive Trade

There’s a thought experiment that was circulating the internet a few months ago: Everyone in the world has to push either a red or a blue button. If the majority of people push the blue button, nobody dies. If the majority of people push the red button, then 50% of blue button pushers die.

Ultimately, the thought experiment interrogates whether we are personally more collectivist or individualistic.

Every trade in FX is kind of like the red button/blue button thought experiment. In this case, though, nobody is even remotely considering the blue button. It’s not that the odds are too high they’d be wrong; it’s that, on the occasions they are wrong, the price for failure is punishing. Red button logic thus rules.

That brings us to an unsettling thought that permeates FX trading: The person on the other side of your trade is actively trying to figure out if you’re dangerous.

Not exactly saloon western dangerous, “dangerous” in the sense that your order might cost them money.

Each time a liquidity provider quotes you a price, they're betting they can hedge the risk of your trade before the market moves against them. Occasionally, the person on the other side of their bet knows something they don’t. Maybe they saw a news wire half a second earlier, or their algorithm detected a pattern. It might even just be that the money-movement in question is very large, its sheer size might push the market in an unfavorable way for the provider.

This mentality fuels what traders call adverse selection, which is when the trades most likely to get filled are also the ones most likely to hurt the person filling them. It's not unique to FX, you’ll find it in every liquid market, but the FX market is fragmented and decentralized. These characteristics make the problem of adverse selection especially prickly.

So liquidity providers are naturally defensive. The mechanisms by which they protect themselves have real consequences for the quality of liquidity you receive, which generally go undisclosed.

Red button resources: Last look, spread widening and flow classification

When providers suspect an order might be harmful to fill, they turn to a few primary tools.

Last look is the most notorious. We've already covered it in-depth; basically, a provider reserves the right to reject your trade even after you've clicked go, within a brief window (typically 10–30 milliseconds) where they can check whether the market has moved. On major platforms, between 18-20% of trades get rejected this way in normal conditions. In volatile periods, that figure climbs.

There are legitimate reasons why last look exists, but the problem isn’t really that it happens; it’s when it does. Providers are more likely to reject a trade moving in your favor. If the market ticks up in that 10-millisecond window and you're buying, the provider declines and you’re obliged to re-enter at a worse price. If, on the other hand, the market ticks against you, your order is happily filled. Over thousands of trades, that asymmetry can accumulate into stiff costs.

Spread widening is subtler but arguably more common. Rather than rejecting outright, a provider simply quotes you a worse price when their systems flag your order as potentially toxic. You still get filled, but at a wider spread than the going market rate. The price might look competitive at first glance. What you don't see is that a cleaner order flow would have gotten a tighter quote.

Flow toxicity scoring is the engine behind both of the above. Providers run constant analysis on the orders coming through their books. They're wondering, How quickly after a price update does this client trade? Do they consistently trade in the direction the market subsequently moves? Do they send very large orders that tend to move the market?

As a result of these largely silent considerations, your order flow could be flagged as toxic, regardless of your intentions. When that happens, you’ll face wider spreads and more rejections. Sometimes you won’t get a quote at all.

None of this is necessarily illegal or even unreasonable. A provider can claim excellent liquidity and tight spreads, and those claims may hold true for small, well-timed and unthreatening orders.

The risk providers manage is real. The issue is, it’s red button thinking. A trade moves too fast for them to wonder whether a well-meaning collectivist is sitting in front of them; it’s in their interest to assume bad faith, so the defenses they erect as a result of this assumption are never disclosed.

In thin markets, basic defenses become knockout punches

If you’re trading in primarily G7 currencies, everything described above is annoying but manageable. Barring a crisis event, those markets are reliably liquid; EUR/USD has sufficient competing providers that if a provider rejects you or widens their spread, alternatives are plentiful. Even a large order may not trigger defensive turkey-dancing.

It’s in thinly-traded markets that the dynamic becomes genuinely punishing. We’re talking exotic currencies, emerging market corridors, any place where only a handful of serious providers exist. These are the reasons why:

1. When you get rejected, there are fewer alternative providers to turn to. You can’t just jump to the next if your USD/NGN trade gets dropped; you’ll probably get kicked back into the same pool of providers, who already know how the others were likely to behave.

2. Exotic pairs are inherently more volatile, so last-look is more likely to produce a price move that triggers rejection. The very conditions that make you need to trade urgently—volatility, a hedge or even just a payment deadline—are the same ones that will make providers more reluctant to fill you.

3. Finally, spread widening in thin markets is harder to detect. In EUR/USD, there are enough available data points for you to have a general sense of what a fair spread looks like. Exotic corridors may not have such clean benchmarks. Providers know this … and in red-button land, it’s only natural for a few to take advantage.

The standard of providing high quality liquidity is easy to meet, most of the time, in the currencies for which our system was designed. (We’ve written a lot about that, too.) But the world is bigger than the G7, which in any case are markets that are very well-served. That’s part of why we set our focus where we do: Beyond that small liquid-rich pool.

Chances are, you’re going to need to move money outside the safe confines of G7 currency eventually. Consider what button your liquidity providers are most likely to push before you have to face the result.