I heard that Karpeles wasn't even running the site when things went down. He'd sold the site to Ross Ulbricht, and then Ulbricht tricked him into taking over Mt. Gox again right before everything hit the fan.

Ah, the "Reverse Karpelès Defense".

Citation needed.

I'm pretty sure it's a joke.

We're not allowed to joke on HN. It's serious business here.

I thought this was a democracy!

Relax, it was a joke.

I have no inside information, but the pattern suggests to me a plausible explanation. And I'm typing on my phone, so I might not explain fully.

Suppose I were wanting to overinflated the value of my business to show to investors or for another reason - I could create an account that had within it a large additional amount of fiat (by entering that value in the DB).

Then, I have an idea... Using that fake deposit, I can start buying BTC. And, as long as the price of BTC is going up, I'm actually printing money and making real fiat out of the initial fake deposit.

This works as long as the price continues to increase and people continue to trade - if either of those factors trends down, the fake fiat will be noticed.

Unwinding this becomes tricky - if you sell too fast, you cause the price decrease. And if anybody gets wind of it, they'll abandon ship.

In some ways, this has an analogue in what happened at Lehman and AIG. The CDO market worked as long as the default assumptions were right and the value of the underlying assets continued to appreciate - as soon as they didn't, the margin requirements wiped out all of their reserve capital.

It also reminds me a bit of QE - the "printed" fiat was inserted in to the market and used to purchase assets in a way that supported the market. When he easing stops, if the value of the assets can't be supported by continuing market pressure, the market for those assets crashes.

If your story is valid then - in some sense - Willy was less risky than Lehman was because the reserve ratio (assuming 10 mn$ of "printed money") was only a fraction of the existing money (I suppose less than 10%), so they operated with a reserve of >=90%

Lehman on the other hand had a leverage of approx. 30 = a reserve of 3.3%

Seems like the existence of a regulator for Lehman / AIG had some impact, just not quite the intended one.

> Unwinding this becomes tricky - if you sell too fast, you cause the price decrease. And if anybody gets wind of it, they'll abandon ship.

My thoughts exactly. There are far less risky ways for an exchange to rip off its customers with out their knowledge, such as front running. That is, the insider can use lag to effectively trade with advance knowledge of the market. (Remember those long bouts of massive lag?) This doesn't require market manipulation and every single pair of trades is profitable.

Another piece of this puzzle might be shenanigans early on in the history of MtGox. There's been some speculation that something happened early on and MtGox lost a lot of BTC (either a hack, a glitch, or possibly a HD failed with a wallet that didn't have a backup). From that point on they basically scrambled to make up the deficit in BTC through various shady means, one of which might be this bot. There's an oblique reference to that theory early on in the report when they mention trading a BTC deficit for a fiat deficit. Basically they traded cash they didn't have for BTC they didn't have but needed in order to prop up the orders they had outstanding.

Most of that was known. This "investigation" doesn't have access to Mt. Gox's internal logs.

The actual police investigation has been unimpressive. The Tokyo Metropolitan Police are quite new at computer crime investigation. Their computer crime unit was established in May 2013, about two months before Mt. Gox started tanking.[1] Still, it's amazing that this case hasn't been cracked yet. It has to be an inside job, and the number of insiders is small.

[1] http://antifraudintl.org/threads/tokyo-police-sets-up-cyber-...

> I only skimmed the article.

Bad idea. You should read the whole article and the linked ones too. There's a lot of information there and it's really important to understand what's going on (as well as understanding how MtGox operated).

> I'm assuming that the bitcoins existed and the trades actually happened -- it would be straight forward to tell from the block chain.

That's a big assumption. You can't tell from the blockchain since trades were done within MtGox's database, while the blockchain was only used for BTC withdrawals.

In fact, AFAIK we can't even tell if the data is real, since this is a leaked log dump and could be altered to frame someone.

3 less legs than he was born with?

So was it Karpeles or one of his employees?

It seems like he would have been the only one in the position to do it (the employees were kept in the dark about a lot of things).

Most employees at most companies have pretty good grasp of things they're officially "kept in the dark" about.

Looking at the graph near the end of the article, I'd be curious to know who (Karpeles or other?) took a family vacation from January 3rd to 14th...

That should be relatively easy for the investigators to find that out.

I think that's what what the term "inside job" suggests.

Yes, but which of "karpeles" or "employee" was it?

I think pan69 parsed the sentence as

    Was the insider in the set {Karpeles, Employee}

Which is true by definition. Instead of

    Which element of the set {Karpeles, Employee} satisfies the condition (==insider)

Which is what zzleeper was intending.

This is why we should implement "xor" in natural language :D Now for the wars on if it's pronounced "zor" or "ex-or".

XOR wouldn't help much in this case. The answer to "Was the culprit A XOR B?" would still be interpreted such that the answer is a Yes or No.

True. It's more like "For condition C, I assert that a single element of set S satisfies C. Which element is it?"

This could be useful for situations like:

Question: "Was it night or day?"

Logician's Answer: "Yes"

Tell me again why we need a new linguistic construction when we have the word "which"?

"So which was it, Karpeles or one of his employees?" "Which was it, night or day?"

I don't understand why a bot is so suspicious. It could have simply been algorithmic trading.[1]

[1]: http://en.wikipedia.org/wiki/Algorithmic_trading

It's suspicious because the money didn't seem to come from anywhere, the BTC didn't seem to go anywhere, and most importantly, it was doing trades while MtGox was offline for everyone else - which implies it was running internally at MtGox, which hints at it being used to artificially drive up the price of BTC.

Thanks, that explains it clearly.