Satoshi Dice -- Statistical Analysis

[Stephen Gornick]

it looks like SatoshiDice is now down 250 BTC

And that's not including the fees it pays on each payout,right?
 46518 X 0.0005 = another 23.259 BTC

[Edit: I overlooked in the OP, fees were included in that 250 BTC total.]

[etotheipi]

it looks like SatoshiDice is now down 250 BTC

And that's not including the fees it pays on each payout,right?
 46518 X 0.0005 = another 23.259 BTC

Actually, the fees are computed along with everything else, and included in that number.  As of my latest post:

Results: 2012-May-13 02:43pm

...

Total Bets Made:               49861
Cumulative Wagers:             19914.74392604 BTC
Cumulative Rewards:            20140.46610080 BTC
Cumulative Fees Paid:             25.06200000 BTC
----
SatoshiDice Profit/Loss:        -250.78417476 BTC

The latest win/lost difference is 225.72 BTC, and then they've paid 25.06 BTC in fees to execute the payouts.   It appears they've paid slightly more than you would expect in fees... which with the above data would be 24.9305, but the script found 25.062 in fees paid.  Not a huge difference, but it means that some transactions must've had more than 0.005 fee.  Perhaps there were a couple winning payouts that required creating huge transactions and required larger fee.  I would have to assume the fee logic is included and automatically adjusted appropriately...

EDIT:  I see you figured it out before I finished posting   Oh well, I'll leave this post here...

[btccomm]

i see that you have a maximum bet per address, but aren't you still susceptible to a http://en.wikipedia.org/wiki/Martingale_%28betting_system%29 martingale betting system?
since you have static 1dice addresses, you could get wiped out pretty quickly by an automated program

[payb.tc]

i see that you have a maximum bet per address, but aren't you still susceptible to a http://en.wikipedia.org/wiki/Martingale_%28betting_system%29 martingale betting system?
since you have static 1dice addresses, you could get wiped out pretty quickly by an automated program

are you talking about this automated program by any chance?

[gmaxwell]

i see that you have a maximum bet per address, but aren't you still susceptible to a http://en.wikipedia.org/wiki/Martingale_%28betting_system%29 martingale betting system?
since you have static 1dice addresses, you could get wiped out pretty quickly by an automated program

Martingale on a negative expectation game still has negative expectation so it's the users who are susceptible to it, not the site.

(Casinos avoid it because its procedurally disruptive, and rapidly ruins customers they're rather milk, not because it makes the games net positive for the players)

[payb.tc]

i see that you have a maximum bet per address, but aren't you still susceptible to a http://en.wikipedia.org/wiki/Martingale_%28betting_system%29 martingale betting system?
since you have static 1dice addresses, you could get wiped out pretty quickly by an automated program

Martingale has negative expectation so it's the users who are susceptible to it, not the site.

(Casinos avoid it because its procedurally disruptive, and rapidly ruins customers they're rather milk, not because it makes the games net positive for the players)

satoshidice has a very wide dynamic range... doing the standard 2x martingale, it allows up to 17 losses in a row... you won't get anywhere near that at a regular casino... more like 4 or 5 in a row.

the key to reducing martingale risk (for the house) is bringing the min/max closer together.

[btccomm]

i see that you have a maximum bet per address, but aren't you still susceptible to a http://en.wikipedia.org/wiki/Martingale_%28betting_system%29 martingale betting system?
since you have static 1dice addresses, you could get wiped out pretty quickly by an automated program

are you talking about this automated program by any chance?

lol, no i didn't see that thread.  but a martingale system seemed a natural response to the satoshidice setup,  you're way ahead of me.  

[gmaxwell]

Next time you update the stats, could you also include "megabytes added to the block chain" and/or "percent of total bitcoin transactions"?

[etotheipi]

Next time you update the stats, could you also include "megabytes added to the block chain" and/or "percent of total bitcoin transactions"?

Updated, and with the calculations you requested.  I don't know why I didn't do it originally, I've been curious myself...

It's actually not as much as I thought.  Only 22% of transactions and transaction bytes are due to SatoshiDice (starting the counters after the first SatoshiDice transaction).

[Stephen Gornick]

UPDATE:  I have looked a little closer and found that the unrewarded wagers are real.  It's likely that many of them are simply wagers that will be returned/rewarded, but hadn't made it into the blockchain yet, as of the moment my script was run.  However there are definitely some that are more than one day old, which suggests that a small fraction of bets are not being returned, at least not in a timely manner.

Hadn't you found that those have since been resolved?

[nguoinhaque]

Next time you update the stats, could you also include "megabytes added to the block chain" and/or "percent of total bitcoin transactions"?

It's actually not as much as I thought.  Only 22% of transactions and transaction bytes are due to SatoshiDice (starting the counters after the first SatoshiDice transaction).

Each bet has 2 transactions. Did you count it?

[etotheipi]

Next time you update the stats, could you also include "megabytes added to the block chain" and/or "percent of total bitcoin transactions"?

It's actually not as much as I thought.  Only 22% of transactions and transaction bytes are due to SatoshiDice (starting the counters after the first SatoshiDice transaction).

Each bet has 2 transactions. Did you count it?

Ack.  Talk about an amateurish mistake.  I'm pretty sure I did not count two transactions.  I counted "wagers". 

Unfortunately, I can't reproduce any of these results any time soon, since my main hard-drive crashed last night...

[apetersson]

Unfortunately, I can't reproduce any of these results any time soon, since my main hard-drive crashed last night...

to restore my faith in sanity of bitcoin users, please tell me you have everything backed up and thus you lost no bitcoins..

[etotheipi]

Unfortunately, I can't reproduce any of these results any time soon, since my main hard-drive crashed last night...

to restore my faith in sanity of bitcoin users, please tell me you have everything backed up and thus you lost no bitcoins..

Paper backups all the way!  And of course, all my Armory code is backed up between multiple systems via github and the dozens of users who have checked out the project...

Interesting update on that: I own a Crucial M4 SSD as my primary drive.  It turns out there's a bug in the default firmware of these drives that causes it to become inoperable after exactly 5,184 hours of up-time.  And yes, I got the drive about 7 months ago.  Just need to upgrade the firmware to get it working again...

So, luckily nothing is actually lost.   Just a day of stress trying to figure out how I'm going to diagnose and resolve this situation. 

[jg]

Wow, what a nasty bug!

I'm glad it wasn't too hard to get it running again!

[etotheipi]

Okay, stats are updated!  And nguoinhaque was right, I wasn't counting two bets.  The actual percentage of blockchain volume is about 45% satoshidice!

[JoelKatz]

satoshidice has a very wide dynamic range... doing the standard 2x martingale, it allows up to 17 losses in a row... you won't get anywhere near that at a regular casino... more like 4 or 5 in a row.

the key to reducing martingale risk (for the house) is bringing the min/max closer together.

Why would you want to reduce risk when you're in it for the long haul because you have infrastructure to pay off and the odds are strictly in your favor? Progressive bettors are *great* for Satoshidice. They drastically increase the total amount wagered and the expected house profit depends heavily on the total amount wagered. Even better, they tend to place larger bets than casual betters, which increases the house's advantage because larger bets are cheaper to pay off compared to the expected house profit.

[jgarzik]

Okay, stats are updated!  And nguoinhaque was right, I wasn't counting two bets.  The actual percentage of blockchain volume is about 45% satoshidice!

Good stuff, thanks for messing around with this.

[apetersson]

it amazes me how much betting with guaranteed losses bitcoin users are willing to make
i see the tx "spam" mostly as ok, it shows that the network can handle it. if the miners think this is too much they will start charging more fees.
what i especially like is that the implied transaction fee including block reward is going way down from 5-6  to about 1.2-1.5 :
http://blockchain.info/charts/cost-per-transaction?daysAverageString=7

[etotheipi]

Statistics updated!  SatoshiDice just hit 100,000 wagers! 

it amazes me how much betting with guaranteed losses bitcoin users are willing to make

Gamblers know the games are designed so the bettor loses money.  But in the short run, the variance of the games is so ludicrously high compared to the house edge, that they can't feel the difference compared to a fair game or even in their favor, even after 1,000 bets.   It's the swings and chance to make some money which is where the fun is.

Smart players who understand it know they're paying for their own entertainment. 
The less knowledgeable players are doing it for the fun of gambling, and probably don't even fully appreciate what 51% house edge means.  They had that one huge winning session once and just want to try to make it happen again.