Bitcoinica model is flawed

[realnowhereman]

I've been thinking hard about how Bitcoinica works, and can only conclude that it's completely flawed.  I hope I'm wrong, so here's my reasoning:

Imagine that Bitcoinica has no reserves.  Mr Short, and Mr Long trade positions with each other at a price of $5 for $50 and 10 BTC respectively.  The trading pool at this point holds $50 and 10 BTC.

Code:

    Price               2.5  3.75    5   7.5    10
    Net long / BTC      0    6.67   10  13.33   15
    Net short / BTC     30  16.67   10   3.33    0
    Net long / dollars  0   25      50  100    150
    Net short / dollars 75  62.5    50  25       0

When the price falls to $2.5, Mr Long is forcibly liquidated; that must be done by trading on Mt.Gox.  Mr Long's 10 BTC (which is still in the pool) gets converted to $25, and stored in the Bitcoinica fund pool.  Mr Short's position at that moment is therefore fully backed, there is $75 in the pool.  What if price continues to drop though?  When price is $1, Mr Short is expecting to get his original $50, plus $50 profit.  The pool only holds $75 though.

The answer is of course, that when Mr Long was liquidated, no trade on Mt.Gox can have happened.  Instead, someone else must be found to take over his position.  That liquidation required a new party to buy 10 BTC at $2.5, leaving the pool with $50 and 20 BTC.  Who is going to take that on though?

Bitcoinica has dealt with this problem by introducing its reserves, and upon liquidation it takes on the counter party position using those reserves.

That is an unsustainable though, those reserves are not infinite.  Travel backward in time, when bitcoins were $0.01 and imagine you had gone long (1:1) with 100 BTC.  That required someone else to go short for $1.  Now BTC are $6 each; and you are worth 200*6 = $1200; at some point that shorter was liquidated, losing their $1; but that leaves us $599 lacking in the fund pool.

Now imagine that same was done at 1:10 leverage, there is no way to magically create the backing for the profits that are due, and Bitcoinica would have had to find (1900*6-1) to cash out the long position.  When the market moves a significant amount, Bitcoinica's reserves will be insufficient to cover the profits that participants are expecting.

Forced liquidations would need to happen in both the negative and positive directions.  When your counter party is liquidated; that is the end of the game for you too, but you end up in profit.

Is no one else suspicious at the admitted losses Bitcoinica made during the huge bounce last week?  If Bitcoinica is working correctly, it should always make money.

[Hunterbunter]

I've been thinking hard about how Bitcoinica works, and can only conclude that it's completely flawed. 

Is no one else suspicious at the admitted losses Bitcoinica made during the huge bounce last week?  If Bitcoinica is working correctly, it should always make money.

Yes, I came to this conclusion too...had an account on there for a couple days to see how it worked, and as soon as I did, took whatever I had off. I don't think there's anything particularly wrong with them taking that risk, since the customer can have the advantage at some point, but from a business perspective, it could cause them problems. People can still go 10:1 long now, max out their reserves without any shorters, pushing up the price forever, beyond even what they would be able to catch up on, and bitcoinica only stands to lose all its reserves to its own customers.

If they were so inclined, the model also adds enormous incentive to deliberately force people out of positions on both ends of the market (so they can re-balance things and profit internally) - pretty much exactly how things happened last week, for example. I'm not  insinuating any wrong doing was done, but if wrong doing were to be done, I believe last week is exactly what it would look like.

[realnowhereman]

Yes, I came to this conclusion too...had an account on there for a couple days to see how it worked, and as soon as I did, took whatever I had off. I don't think there's anything particularly wrong with them taking that risk, since the customer can have the advantage at some point, but from a business perspective, it could cause them problems. People can still go 10:1 long now, max out their reserves without any shorters, pushing up the price forever, beyond even what they would be able to catch up on, and bitcoinica only stands to lose all its reserves to its own customers.

I think it's worse than that.  When their customers make too much money (as they will if bitcoins go up while "*" indicates that Bitcoinica is maxed out long).  The short's losses can only cover those profits up to their deposits; after that there is reported profit being made by the longs that is not really being generated.  The long accounts will all show huge profits, but there will be no real money to cover those profits.  Customers will surely realise that at some point, and effectively a run will ensue.  Whoever gets their money out first will profit.

So... it's not simply that bitcoinica will lose money to its customers; it's that there will not be enough money to pay winners.  The customers are not making the profits they think they are.  However, they are taking the risks they think they are, and some risks that they don't know about.

[teek]

Interesting thread, subscribing.

+1

[Clark]

Has anyone found any resources on how to start a brokerage? Or a retail forex shop?

This is essentially what Bitcoinica is.

[zhoutong]

OP made a fundamental mistake by assuming there are no reserves. In normal situation (without asterisk), Bitcoinica's reserves are able to cover all open positions. That means, Bitcoin price can go to $10,000 or $0.01 and we don't have to take any market risk out of this.

When there is asterisk, we can still cover all existing positions, but we can't cover more. That's why you can always reduce your positions without worrying too much.

I have run through many simulated market moves before I release the algorithms to the public, and our financial situations are much better than expected.

[Dutch Merganser]

you cant go 10:1 over 2000 btc just saying

Assuming using only one account you are correct.

Bitcoinica just locks the front door when things get away from them, there are pluses and minuses to that approach.

[SgtSpike]

OP made a fundamental mistake by assuming there are no reserves. In normal situation (without asterisk), Bitcoinica's reserves are able to cover all open positions. That means, Bitcoin price can go to $10,000 or $0.01 and we don't have to take any market risk out of this.

When there is asterisk, we can still cover all existing positions, but we can't cover more. That's why you can always reduce your positions without worrying too much.

I have run through many simulated market moves before I release the algorithms to the public, and our financial situations are much better than expected.

Why did you lose money?  Just because the price swings were so violent that you weren't able to liquidate positions at the proper price points?

[Qoheleth]

The answer is of course, that when Mr Long was liquidated, no trade on Mt.Gox can have happened.  Instead, someone else must be found to take over his position.  That liquidation required a new party to buy 10 BTC at $2.5, leaving the pool with $50 and 20 BTC.  Who is going to take that on though?

Bitcoinica has dealt with this problem by introducing its reserves, and upon liquidation it takes on the counter party position using those reserves.

I think this is the mistake in your reasoning.

When Mr. Short becomes unmatched by a Mr. Long, Bitcoinica doesn't step into the shoes of Mr. Long. Instead, Bitcoinica uses the BTC in its reserve to physically sell the BTC that Mr. Short has "virtually sold" by taking a short position; the new Mr. Long is someone on Mt. Gox who physically bought those coins.

That's the point of the reserves - when the shorts and longs don't balance internally, it's used to translate them to actual buys and sells on the open market.

So things look like this:
Mr. Long buys 10BTC for $50
Mr. Short buys -10BTC for -$50

At 1BTC=$2.50:
Mr. Long is liquidated for $25.
Bitcoinica sells 10BTC from its reserve on Mt. Gox for $25.

At 1BTC=$1:
Mr. Short sells -10BTC for -$10
Mr. Short gains $50 (the price where he opened the position) - $10 (the price at which he closed the position) = $40.
Bitcoinica buys 10BTC for $10.

The numbers balance.

That is an unsustainable though, those reserves are not infinite.  Travel backward in time, when bitcoins were $0.01 and imagine you had gone long (1:1) with 100 BTC.  That required someone else to go short for $1.  Now BTC are $6 each; and you are worth 200*6 = $1200; at some point that shorter was liquidated, losing their $1; but that leaves us $599 lacking in the fund pool.

Mr. Long buys 100 BTC for $10.
There's no Mr. Short! So Bitcoinica backs this by going to Mt. Gox and physically buying 100BTC with $10 from its reserve.
A year later, Mr. Long cashes out his position of 100 BTC for $6 each - for a total of $600 minus the $10 from the original long position.
Bitcoinica sells those physical 100BTC for $600, passing $590 to Mr. Long and putting the remaining $10 back into its reserve.

The numbers balance.

The only problem here is that Bitcoinica doesn't have infinite funds to lend out. Their funds aren't at risk; when all positions liquidate, the reserve is just as big as it was when there were no positions at all. They just don't have the capital to materialize an unlimited number of positions.

[Qoheleth]

(I think I should add that there are organizations that take the opposite position when the shorts and longs don't balance. Such operations are called "bucket shops" and it's against the law in many jurisdictions.)

[realnowhereman]

OP made a fundamental mistake by assuming there are no reserves. In normal situation (without asterisk), Bitcoinica's reserves are able to cover all open positions. That means, Bitcoin price can go to $10,000 or $0.01 and we don't have to take any market risk out of this.

If you finished reading my post you will know that while I started with the no reserve case; I didn't finish with it.

Okay; here's another situation.  Bitcoin is at $1; and I go long for 100 BTC.  There is no one to short against me, so Bitcoinica's reserves are used instead.  Your only option is to take the opposing position; i.e. to take on the short that no one else will.  At 1:1, I'll have 200 BTC of value, with a $100 debt still owed to Bitcoinica's reserves.

Now, price goes to $10,000 as you suggest.  My 200 BTC is worth $2,000,000; and I owe $100.  Bitcoinica's short can at most have supplied me $100 of that $2,000,000 balance.

While you might reasonably say, "well Bitcoinica has reserves to cover that", it does not matter -- you have lost on that trade.  If you can lose on trades then you are a speculator not a broker.  Eventually the market will move against you and you won't be able to cover your bets.

Now... my actual point was that opposing longs and shorts only balance while the losing side has a positive net balance.  When one side is liquidated, then there is no source for the other side's profit.  Either you find someone else to take over that (now manifestly losing) opposing position or Bitcoinica starts supplying the profit.

When there is asterisk, we can still cover all existing positions, but we can't cover more. That's why you can always reduce your positions without worrying too much.

Nonsense.  You might be able to cover at the current price, but if the market moves, you will eventually run out of money to supply the unlimited profit potential profit of your customers.  It's certainly not supplied by the losses of another customer as those losses are limited to their deposit; whereas profits are unlimited.

Your only option would be to actually buy every customers position.  So if I go long at 100 BTC 1:1; you have to buy another 100 BTC.  But then if someone goes short for $100 you have to find another $100 to match them (it is not sufficient to sell BTC to cover it; since you would then not have enough BTC to cover the long).

I have run through many simulated market moves before I release the algorithms to the public, and our financial situations are much better than expected.

That's lovely for you.  But I don't believe it.  Going long at say 1:10 with 1000 BTC when bitcoins go from $1 to $1000 would require you to find $1,000,000 for me to cash out.  The opposing shorter of $1000 (at $1 per BTC) can at most supply $1000.  What if 100 people do the same thing?  1000?  Bitcoinica cannot possibly have reserves to cover the profit on those positions given that, contrary to what everyone seems to think, the opposing customers cannot be covering it.

[realnowhereman]

The answer is of course, that when Mr Long was liquidated, no trade on Mt.Gox can have happened.  Instead, someone else must be found to take over his position.  That liquidation required a new party to buy 10 BTC at $2.5, leaving the pool with $50 and 20 BTC.  Who is going to take that on though?

Bitcoinica has dealt with this problem by introducing its reserves, and upon liquidation it takes on the counter party position using those reserves.

I think this is the mistake in your reasoning.

When Mr. Short becomes unmatched by a Mr. Long, Bitcoinica doesn't step into the shoes of Mr. Long. Instead, Bitcoinica uses the BTC in its reserve to physically sell the BTC that Mr. Short has "virtually sold" by taking a short position; the new Mr. Long is someone on Mt. Gox who physically bought those coins.

That's the point of the reserves - when the shorts and longs don't balance internally, it's used to translate them to actual buys and sells on the open market.

Well, I demonstrated that even in the case of balanced longs and shorts, there is still danger.  When the losing side is liquidated, the winning side is not covered for any further profit.

If I go long for 100 BTC at $1, then a short for $100 will cover $100 of my profit if prices go up.  What happens when that $100 is used up?  The only possible source to cover more increases in my long profit is to buy enough BTC for my long... i.e. another 100 BTC.

So things look like this:
Mr. Long buys 10BTC for $50
Mr. Short buys -10BTC for -$50

At 1BTC=$2.50:
Mr. Long is liquidated for $25.
Bitcoinica sells 10BTC from its reserve on Mt. Gox for $25.

At 1BTC=$1:
Mr. Short sells -10BTC for -$10
Mr. Short gains $50 (the price where he opened the position) - $10 (the price at which he closed the position) = $40.
Bitcoinica buys 10BTC for $10.

The numbers balance.

Yes; the numbers will always balance.  The problem is that as the opposing side passes to a negative net position and is liquidated, it is only the numbers that balance.  They start to require the ability to trade negative amounts on Mt.Gox -- which is obviously impossible.  The problem is the further profits after one side is liquidated.

That is an unsustainable though, those reserves are not infinite.  Travel backward in time, when bitcoins were $0.01 and imagine you had gone long (1:1) with 100 BTC.  That required someone else to go short for $1.  Now BTC are $6 each; and you are worth 200*6 = $1200; at some point that shorter was liquidated, losing their $1; but that leaves us $599 lacking in the fund pool.

Mr. Long buys 100 BTC for $10.
There's no Mr. Short! So Bitcoinica backs this by going to Mt. Gox and physically buying 100BTC with $10 from its reserve.
A year later, Mr. Long cashes out his position of 100 BTC for $6 each - for a total of $600 minus the $10 from the original long position.
Bitcoinica sells those physical 100BTC for $600, passing $590 to Mr. Long and putting the remaining $10 back into its reserve.

The numbers balance.

This is the very definition of the problem.  Bitcoinica has to do exactly this for every long and every short.  The imbalance is not the only thing that needs backing, every position needs backing as the shorts only cover the longs to their margin; and the profit for the long can exceed that margin.

The only problem here is that Bitcoinica doesn't have infinite funds to lend out. Their funds aren't at risk; when all positions liquidate, the reserve is just as big as it was when there were no positions at all. They just don't have the capital to materialize an unlimited number of positions.

I don't think their funds are at risk when people liquidate.  The liquidates are easy.  The problem is the profits.

[rjk]

So, dump $2 million on the market after creating several Bitcoinica accounts - some long, some short, and watch how it reacts. I'd be interested on the outcome, especially since the price would jump up and I would be rich!

[realnowhereman]

So, dump $2 million on the market after creating several Bitcoinica accounts - some long, some short, and watch how it reacts. I'd be interested on the outcome, especially since the price would jump up and I would be rich!

Yeah, I'll just "dump $2 million".

Hahaha.  I think you've mistaken me for a millionaire.

[SgtSpike]

Your only option would be to actually buy every customers position.  So if I go long at 100 BTC 1:1; you have to buy another 100 BTC.  But then if someone goes short for $100 you have to find another $100 to match them (it is not sufficient to sell BTC to cover it; since you would then not have enough BTC to cover the long).

My understanding is that is exactly what is happening.  Why would you think otherwise?

[realnowhereman]

Your only option would be to actually buy every customers position.  So if I go long at 100 BTC 1:1; you have to buy another 100 BTC.  But then if someone goes short for $100 you have to find another $100 to match them (it is not sufficient to sell BTC to cover it; since you would then not have enough BTC to cover the long).

My understanding is that is exactly what is happening.  Why would you think otherwise?

People seem to think that Bitcoinica are only using their reserves for the imbalance in shorts and longs; my argument is that shorts do not fully balance longs.  Therefore the reserves have to cover every position, not just the imbalance.

[notme]

OP made a fundamental mistake by assuming there are no reserves. In normal situation (without asterisk), Bitcoinica's reserves are able to cover all open positions. That means, Bitcoin price can go to $10,000 or $0.01 and we don't have to take any market risk out of this.

When there is asterisk, we can still cover all existing positions, but we can't cover more. That's why you can always reduce your positions without worrying too much.

I have run through many simulated market moves before I release the algorithms to the public, and our financial situations are much better than expected.

Why did you lose money?  Just because the price swings were so violent that you weren't able to liquidate positions at the proper price points?

This... the term is slippage.  He couldn't liquidate fast enough.

Anyone else think it's odd Zhoutong's reply is essentially ignored?  Why pay attention to facts when you can guess about how things work?

[realnowhereman]

OP made a fundamental mistake by assuming there are no reserves. In normal situation (without asterisk), Bitcoinica's reserves are able to cover all open positions. That means, Bitcoin price can go to $10,000 or $0.01 and we don't have to take any market risk out of this.

When there is asterisk, we can still cover all existing positions, but we can't cover more. That's why you can always reduce your positions without worrying too much.

I have run through many simulated market moves before I release the algorithms to the public, and our financial situations are much better than expected.

Why did you lose money?  Just because the price swings were so violent that you weren't able to liquidate positions at the proper price points?

This... the term is slippage.  He couldn't liquidate fast enough.

Anyone else think it's odd Zhoutong's reply is essentially ignored?  Why pay attention to facts when you can guess about how things work?

Given that you know what the liquidation threshold is well in advance, why is "speed" a factor?  You could have the order sat on Mt.Gox's books from the moment the position is established.  The market then couldn't move to a "slippage" price without passing through your order.

[SgtSpike]

OP made a fundamental mistake by assuming there are no reserves. In normal situation (without asterisk), Bitcoinica's reserves are able to cover all open positions. That means, Bitcoin price can go to $10,000 or $0.01 and we don't have to take any market risk out of this.

When there is asterisk, we can still cover all existing positions, but we can't cover more. That's why you can always reduce your positions without worrying too much.

I have run through many simulated market moves before I release the algorithms to the public, and our financial situations are much better than expected.

Why did you lose money?  Just because the price swings were so violent that you weren't able to liquidate positions at the proper price points?

This... the term is slippage.  He couldn't liquidate fast enough.

Anyone else think it's odd Zhoutong's reply is essentially ignored?  Why pay attention to facts when you can guess about how things work?

Given that you know what the liquidation threshold is well in advance, why is "speed" a factor?  You could have the order sat on Mt.Gox's books from the moment the position is established.  The market then couldn't move to a "slippage" price without passing through your order.

Let's do a very basic scenario.

Someone is long with a base of $5 and liquidation at $4.

MtGox has open buy orders of $4.50 and $3.50.

The $4.50 buys are filled, and the "price" suddenly drops to $3.50, because there is nothing at $4.00.

The person gone long is liquidated, but zhout has no choice but to liquidate the Bitcoins at $3.50 instead of $4.00, because that is the going market price.

Normally, the spread between bids is very small, so $4 would be liquidated at $3.995 or something, and zhout makes the money off the maintenance fee (if the liquidation is $4.00 to the user, then the "true" liquidation is actually $3.75 or $3.90 or something).  But with large price movements and lots of liquidations in a short period of time, not enough buyers can be found on MtGox to sell all of the liquidations at their liquidation prices, and they slide the price even further beyond the maintenance fee, causing a loss.

Anyone feel free to correct me if I am wrong - this is all just guessing based on the way I would set it up.

[notme]

Given that you know what the liquidation threshold is well in advance, why is "speed" a factor?  You could have the order sat on Mt.Gox's books from the moment the position is established.  The market then couldn't move to a "slippage" price without passing through your order.

Let's do a very basic scenario.

Someone is long with a base of $5 and liquidation at $4.

MtGox has open buy orders of $4.50 and $3.50.

The $4.50 buys are filled, and the "price" suddenly drops to $3.50, because there is nothing at $4.00.

The person gone long is liquidated, but zhout has no choice but to liquidate the Bitcoins at $3.50 instead of $4.00, because that is the going market price.

Normally, the spread between bids is very small, so $4 would be liquidated at $3.995 or something, and zhout makes the money off the maintenance fee (if the liquidation is $4.00 to the user, then the "true" liquidation is actually $3.75 or $3.90 or something).  But with large price movements and lots of liquidations in a short period of time, not enough buyers can be found on MtGox to sell all of the liquidations at their liquidation prices, and they slide the price even further beyond the maintenance fee, causing a loss.

Anyone feel free to correct me if I am wrong - this is all just guessing based on the way I would set it up.

I believe you are correct SgtSpike.  realnowhereman is also correct that he could avoid slippage by putting orders in place.  This has it's own issues though since if the user closes or changes their position, the order would need to be updated.  If the market took the order before it could be adjusted, again Zhoutong has loss.