SatoshiDICE.com - The World's Most Popular Bitcoin Game

[deepceleron]

satoshi dice isn't paying fees for returning below minimum transactions anymore. but these transactions aren't being relayed as they don't met the free transactions conditions (https://en.bitcoin.it/wiki/Transaction_fees). so, what will happen with these coins?

example:

http://blockchain.info/tx-index/43841974/2dc3e903d6c30fffb8c606ed4dd26d57a3c5436128dd8267c0379467681156d3

The transaction was relayed - that is how blockchain.info knows about it - it was broadcast around the network. However, as it didn't include the minimum 0.0001 per KB fee for relay through the p2p network, only those that have a direct IP connection to Satoshidice's server or another peer that has relaxed relaying policies would know about it, and the chance that there is a miner with this transaction waiting in their memory pool is low.

Bitcoins don't magically end up back in the wallet if they are not included in a block. The Bitcoin wallet software considers spent coins as spent, and it takes manually removing the transaction through wallet editing tools to restore the balance shown in the wallet - likely SatoshiDice is not going to do this for 0.001 BTC that they consider "sent" (and should have just kept/ignored) so the coins will probably be considered lost.

As a followup, this transaction was included in Block 216642 by Eclipse mining pool after 10 hours, along with several other transactions that did not pay minimum fees.

[1680214080]

1680214080

[evoorhees]

It looks like S.Dice now waits for 1 confirmation before handing out BTC. Is this for real?

It might wait for a confirmation if the transaction looks really suspicious, but in the vast majority of bets no confirmation is required.

[Tomatocage]

Is there a historical chart somewhere of all the random numbers rolled since day 1?

[evoorhees]

Is there a historical chart somewhere of all the random numbers rolled since day 1?

I don't believe so

[dooglus]

Is there a historical chart somewhere of all the random numbers rolled since day 1?

I don't believe so

Several months ago I calculated a rough distribution of the random numbers.  It's rough because I can't be sure which day's secret was used for each bet.  Some bets are processed long before they get into a block, and some long after.  But if I use the timestamp on each bet's block, I get the correct date more often than not.

I just recalculated the distribution again and plotted it.

Here's a plot of the number of occurrences of each random number.  The most common number is 45686, appearing 66 times, and the least common is 20190, appearing only 12 times:



Here's a plot of the percentage of random numbers less than each given number.  It should be a straight line, and pretty much is:



Here's a list showing, for each satoshidice bet, how many random numbers were below the bet threshold (A), how many we would expect to be under that threshold (B), and A as a percentage of B.

For example, since there were 2258204 lucky numbers generated, we would expect 2258204/65536 = 34.46 of the random numbers to be less than 1.  In fact, only 29 were below 1, which is 84.16% of the expected amount.

bet     actual   expected percentage
---     ------   -------- ----------
<1          29      34.46  84.16%
<2          66      68.91  95.77%
<4         122     137.83  88.51%
<8         256     275.66  92.87%
<16        541     551.32  98.13%
<32       1107    1102.64 100.40%
<64       2174    2205.28  98.58%
<128      4431    4410.55 100.46%
<256      8821    8821.11 100.00%
<512     17640   17642.22  99.99%
<1000    34625   34457.46 100.49%
<1500    51827   51686.19 100.27%
<2000    68942   68914.92 100.04%
<3000   103308  103372.38  99.94%
<4000   137852  137829.83 100.02%
<6000   206324  206744.75  99.80%
<8000   275210  275659.67  99.84%
<12000  413822  413489.50 100.08%
<16000  551369  551319.34 100.01%
<24000  828042  826979.00 100.13%
<32000 1104104 1102638.67 100.13%
<32768 1130370 1129102.00 100.11%
<48000 1654832 1653958.01 100.05%
<52000 1792371 1791787.84 100.03%
<56000 1929924 1929617.68 100.02%
<60000 2067668 2067447.51 100.01%
<64000 2205597 2205277.34 100.01%

In summary, it looks like the random numbers are pretty evenly distributed in the range 0-65535.

[zvs]

How come there aren't like... five million copycat sites of this yet?  All I can think of is BTCDice and it doesn't even seem to work?

[zvs]

<8000   275210  275659.67  99.84%
<12000  413822  413489.50 100.08%

no wonder i always lose.  damn under 8000 bets.  i  sense foul chicanery at work

[organofcorti]

<8000   275210  275659.67  99.84%
<12000  413822  413489.50 100.08%

no wonder i always lose.  damn under 8000 bets.  i  sense foul chicanery at work

You must be a 0.1%er

In seriousness though, I posted the CDF probabiities for the winning bets a while back and even though the < 16 bets seem to be significantly below expected, they're probably (haven't worked it out) still inside a reasonable confidence zone since there's so few expected wins yet.

[Herbert]

How come there aren't like... five million copycat sites of this yet?  All I can think of is BTCDice and it doesn't even seem to work?

I think because it actually is harder to get it right than it seems. You need a deep understanding of the bitcoin protocol as most of the stuff you have to do is beyond the capabilites of the standard client. Besides BTCDice if I remember correctly I think there were one or two additional 1:1 copies that also went down quickly. And SDice will have forever the first-mover advantage, so any 1:1 copy will have a very hard time.

If you want to compete with SDice you have to differentiate, not just copy (which is what I am trying with bitbattle.me ).

[kano]

How come there aren't like... five million copycat sites of this yet?  All I can think of is BTCDice and it doesn't even seem to work?

I think because it actually is harder to get it right than it seems. You need a deep understanding of the bitcoin protocol as most of the stuff you have to do is beyond the capabilites of the standard client. Besides BTCDice if I remember correctly I think there were one or two additional 1:1 copies that also went down quickly. And SDice will have forever the first-mover advantage, so any 1:1 copy will have a very hard time.

If you want to compete with SDice you have to differentiate, not just copy (which is what I am trying with botbittle.me ).

... and then latch on to SatoshiDice and post ads about your site in here and pretend you didn't ...

[MPOE-PR]

How come there aren't like... five million copycat sites of this yet?  All I can think of is BTCDice and it doesn't even seem to work?

Bankroll - how much does an upstart site need to hold off someones martingale? How about multiple whales playing at the same time?

SD (and competitors) need to be able to go out thousands of coin while waiting for the player to bust.

This is so sweet. Back in August people were all liek

Sorry, but for a new unproven (yes 3 months is unproven) business with an easily reproducible model, where the margins will come under pressure from competitors, and which will inevitably succumb to the "It was fun but I've gotten bored of that" novelty value fatigue, I think you should be factoring in considerable reduction, not growth.

Meanwhile the competition is liek

Hmm, it looks like we were overly optimistic regarding how many people would use the site. In light of this we are considering closing the site down. If anyone is interested in purchasing the script and taking over please PM me.

These coincidences. Damn them!

[deepceleron]

How come there aren't like... five million copycat sites of this yet?  All I can think of is BTCDice and it doesn't even seem to work?

Bankroll - how much does an upstart site need to hold off someones martingale? How about multiple whales playing at the same time?

SD (and competitors) need to be able to go out thousands of coin while waiting for the player to bust.

A "Martingale" is a failure betting fallacy, and a casino can only hope for players so stupid. It will result in a player going bust far quicker than the casino. If the bet unit is 1 at 50% odds, and the betting can continue for many rounds of doubling (say 10 rounds before hitting a limit - which is 1024x) then there is little risk to the Casino. Even a win with this "strategy" means the casino wins 1+2+4+8+16+32+64+128+256+512, and if the player wins the final bet the player's total winnings is 1 bet unit; however the player has statistically paid more in house edge when betting those amounts then the total winnings; the bets are independent everywhere except in the player's mind.

Google "risk of ruin". The profit/loss of a casino over the long term is a random walk, and there is always a very small chance that any casino will go bust. If I have a 1 BTC coin flip game and have 5 BTC in the bank, I can let a player play four times with zero risk; there is a one-in-32 chance that the player will bust me in five flips. If I let the player play an infinite amount of time, there is a 100% chance the player will bust me (at some point in the future, maybe after the sun explodes). With a house edge, I can make the risk approach zero within a normal timeline.




If I open a casino tomorrow, there is an extremely small chance that every bet ever placed at the casino will be won by the player. One would need to adjust the gaming rules and house edge to minimize the chance of casino ruin to something you consider safe - do you want a 1% of going bust; a .001% chance of going bust, etc? SD has minimized this risk while having a bankroll to pay out some very high rewards on the higher odds games.


Answering the original question, "why aren't there five million copycat Googles, eBays, Amazons?" If you do something first, and do it as good as any competitor might, there is market inertia that presents a barrier to entry for others.

[organofcorti]

How come there aren't like... five million copycat sites of this yet?  All I can think of is BTCDice and it doesn't even seem to work?

Bankroll - how much does an upstart site need to hold off someones martingale? How about multiple whales playing at the same time?

SD (and competitors) need to be able to go out thousands of coin while waiting for the player to bust.

A "Martingale" is a failure betting fallacy, and a casino can only hope for players so stupid.

Mostly true, especially if it's a naive "double up" with each loss. I derived a martingale strategy that can be used to significantly increase the number of bets before bankruptcy [strategy explained here]. The end result is the same as using a naive strategy, but the "entertainment" value is greater.

[dooglus]

the < 16 bets seem to be significantly below expected

Note that the stats I posted were generated by looking at all 2 million random numbers generated so far and working out how many we would expect to be below 16, then counting how many were actually below 16.  I didn't just look at the random numbers generated for the "lessthan 16" bet, but all random numbers for all types of bet.

[nimda]

<8000   275210  275659.67  99.84%
<12000  413822  413489.50 100.08%

no wonder i always lose.  damn under 8000 bets.  i  sense foul chicanery at work

You must be a 0.1%er

In seriousness though, I posted the CDF probabiities for the winning bets a while back and even though the < 16 bets seem to be significantly below expected, they're probably (haven't worked it out) still inside a reasonable confidence zone since there's so few expected wins yet.

Somewhere many pages back I used binomial distribution to calculate the "confidence zone;" nothing was off enough to suspect foul play.
http://en.wikipedia.org/wiki/Binomial_distribution
I imagine most people here should be able to follow it; the most difficult math used is factorials.

[michaelmclees]

I wonder if there could be a Satoshi Dice game where the casino destroys a portion of the coins it earns, thus increasing the $ value of the coins in its pocket.

[bowen151]

That would mean the casino is essentially gambling with its own funds in my eyes. It would have to hope that the $ value goes up enough to cover the amount that was "destroyed" in the first place, you also would have problems with transparency and showing people how you are "destroying" the coins and not just keeping them.

[Matthew N. Wright]

That would mean the casino is essentially gambling with its own funds in my eyes. It would have to hope that the $ value goes up enough to cover the amount that was "destroyed" in the first place, you also would have problems with transparency and showing people how you are "destroying" the coins and not just keeping them.

Could be as easy as publicly sending to a black hole address. This is a silly idea just the same.

[dooglus]

I wonder if there could be a Satoshi Dice game where the casino destroys a portion of the coins it earns, thus increasing the $ value of the coins in its pocket.

Destroying part of your assets never increases your net worth.

you also would have problems with transparency and showing people how you are "destroying" the coins and not just keeping them.

There's no private key for public key 0, so sending coins directly to that public key would mean they were destroyed.

[Tomatocage]

A "Martingale" is a failure betting fallacy, and a casino can only hope for players so stupid. It will result in a player going bust far quicker than the casino.

I can believe this.  I played over a thousand rolls this past week alone in the 50% bracket. I always start out with .05 BTC, and I've gone on  7x losing streaks (Lose .05 -> .10 -> .20 -> .40 -> .80 -> 1.60 -> 3.20) about a half dozen times. However, after looking at my trade history, I've NEVER gone beyond 5 straight wins, ever. 50% my ass.