But what does "intention to cancel" mean when you're offering to sell stock at an above-market price? What if I'm perfectly happy letting someone buy that stock from me because I can replace it at a lower price once the market settles? I'm still trying to manipulate the market prices, but am I "spoofing"?

If you're actually willing to trade at that above-market price if the market moves suddenly, then that's a legitimate strategy and not manipulation. You're playing the role of a shock-absorber. If your plan is to cancel as quickly as possible if you see the market moving toward you, then that's illegal.

But I can do both of those at the same time. Let's say I'm willing to trade at that price if the market moves suddenly, but I want to change my offer if the market moves slowly. Am I intending to cancel now?

he's referring to the unconditional intention to cancel an order. if any of the future states of the world result in you wanting to trade the order that you placed, it is a valid order.

Great, then the smart spoofer isn't spoofing! Because they accept that they might actually trade the order.

Even if they admit to spoofing as a general activity, they're just talking about the common case. They're not placing orders with unconditional intention to cancel.

Well that doesn't lead to a very helpful law...

You could say the same thing about almost any law. The difference between an accidental fire and arson is criminal intent. While in theory it's not possible to observe someone's inner mental state, in practice their outward actions usually give a pretty good clue, and edge cases where the same actions could be interpreted either way are rare. The same is true in spoofing: while it's theoretically possible that exactly the same strategy might be used by a legitimate trader and by a spoofer, differing only by intent, in practice some strategies are overwhelmingly more likely to be spoofing. The edge cases are tricky, but they don't invalidate the entire concept.

In this particular case, the trader in question would have been financially ruined if his spoofed orders had been executed, so it's pretty clear he didn't intend for them to be.

Caveat: In my opinion, this isn't a distinction well suited to the criminal justice system. I'd rather that exchanges identify probable spoofers and ban them from trading (after some warnings), with the burden of proof (in arbitration) on the accused to show that the suspicious orders were part of a legitimate strategy.

>In this particular case, the trader in question would have been financially ruined if his spoofed orders had been executed, so it's pretty clear he didn't intend for them to be.

Are you sure? If he bought the stocks at the heightened price he could just resell them right away. I'd have expected it would take a much more permanent shift to ruin him.

Or if I put it another way: It seems to me that there is a solid method for making money that's 95% the same as spoofing and 5% actually being happy when your "fake" orders execute. So the intent is still to manipulate the market but only through truthful orders. Then you can no longer look at this pattern of behavior and say that it's overwhelmingly likely to be spoofing. Now what do you do? Do any laws stop this?

Which makes the law entirely unenforceable unless the perpetrator is dumb / arrogant / ignorant enough to brag about it...

But these are high-frequency trades, which mean that they're executed by a computer programmed to do that. If you can examine exactly what it was programmed to do then that could probably help prove intent.

"You're honor, it wasn't me, it just deep-learned it that way."

If a program uses AI to come up with strategies, how do you make sure it doesn't learn to spoof? Coming up with the goal-formulation of "don't spoof" seems like it would require the program to introspect about its own intent. Dunno, I would like to hear what kind of metrics one would use to make the learning algorithm reject spoofing strategies. I guess it would be something similar to whatever heuristics the SEC uses to flag potential spoofing, but they get a lot of false positives and almost never convict someone on metrics.