Yes, you're right. It should be more visible, or people won't even know about it.


Yes. The lottery could have a house edge greater than 1%, with the extra going to fund the jackpot.


1) Investors don't want bigger bets, they want lots of small bets (assuming the total volume is the same). Bigger bets means bigger variance. Lots of small bets summing to the same amount gives the same expectation, but with a much lower variance. When the odds are in your favour, you don't want too much variance.

2) Tiger suggested having the "points" given out according to the size of the bet. That seems reasonable. Someone hitting the jackpot number while making a tiny bet will get a tiny number of points. That's a good point though - what if only one person hits the 99.9999 or whatever one day, and does it betting just 1 satoshi? Even though he only gets a tiny number of points, he still has all the points awarded that day. Does he still get the full 1% daily payout? If the pot is 1000 CLAMs, that's a 10 CLAM payout for a 1 satoshi bet...


Thanks for them.

Thanks for taking the time.


100% of my stakes? I don't like that idea one little bit! Do you mean the jackpot account's stakes go to the jackpot account? That seems reasonable.


When you say "players", you mean "investors", right? Because they're the only ones who benefit from staking. Note that most weeks, for most investors, their return from staking is bigger than their return from players losing. You can see this from the fact that the site has staked 200k, but only won 100k from players - staking is twice as big as gambling profits. So it sounds like you're asking investors to give up 33% of their expected income if I've understood correctly.


I'd like to have 00.0000 count as well. I don't want to bias the jackpot in favour of people who only like to bet 'hi'. I'd also make it that your bet has to be a winning bet to qualify, so it's still only 1 in a million that counts. If someone rolls 99.9999 while betting 'lo', the site could comment on the fact in chat, but not award any thing. Or maybe award 10% of the points they would have got if they rolled correctly, as a booby prize.


I think it will grow towards a limit, and pretty much stay there. We can consider the inflow to be roughly constant, and the outflow to be 17% per week (seven 1% daily payout, and one 10% weekly payout). So the pot will reach a point where the weekly inflow is equal to 17% of the pot. Or in other words the pot will grow to be 6 times the size of the weekly inflow.


Again, I'm guessing you mean 'investors' when you say 'players' here. I think it's good to make some kind of incentive for investors to contribute to the pot, but I don't think this works. If I'm an investor with 10k invested, I'm making quite a lot from staking. I don't want to give half of it up just to get double jackpot points. Instead I can move 10 CLAMs to a new account, invest it, and set just that account to give up half its staking rewards. Then bet on that account and get double jackpot points. In doing so I've managed to keep almost all my staking rewards and still get double jackpot points.


I don't understand. Whether or not the points are reset each week, the weekly payout is always going to be 10% of the pot as I understood the proposal. Taking 10% out each week will limit its growth, of course - but I don't see how the points affect it.


I don't think we can make things +EV for everyone. Somebody has to pay for the jackpot. Adding a jackpot is increasing the player's expectation, and so it has to reduce someone else's expectation. But I agree that asking investors to give up half their staking income for improved jackpot payoffs is unlikely to be well received.

It seems the bittrex wallet got so full of dust outputs that it was unable to make any outgoing payments without exceeding the transaction size limit.

I made a change to my CLAM repository which aims to prevent this from happening again in the future. It could do with some peer review and testing before it gets into the official releases.

The site profit went over 100k and stayed there for 6 hours or so, until:



Here's the recent profit chart:

You're missing an important one: both the betid and the time need to be prime. That's about a 1 in 440 chance if the site is fair, and worse if they decide to cheat. So it's a 1-in-4.4 million chance of hitting the jackpot per bet. But if you bet 1 satoshi each time, the prize is 0.2 BTC (20 million satoshi) and so it's very +EV for you - in in 4.4 million chance of hitting 20 million times payout... Mr. DaDice Staff is apparently working on a addressing this concern for me.

As for your "how does it charge?" question, I don't think that has been answered anywhere.

It could be an option in the account tab, to opt in or out of the jackpot. Most people would leave that at its default, so we would need to decide what the default is. If the default is to opt out, players get the 1% edge they're used to, but almost nobody contributes to the pot. And if the default is to opt in, players will be confused as to why their 49.5% bets no longer pay a full 2x.


I'm thinking to have the jackpot amount stake. It could be in a separate address, and just get the full 100% of any staking rewards that that address earns. Staking returns are bigger than investment returns, and are always positive. I don't think it's good for the size of the jackpot to go down when players win.


I don't want to take anything from the investors that they didn't already sign up for. They agreed to 10% commission, so that's what I take.


That's a decent idea. Maybe a small daily and bigger weekly draw, where every bet has a chance of winning in proportion to the size of the bet.

Warning: stream of consciousness rambling incoming. Skip if you like - just trying to figure out how it could work...

It's not immediately clear how to make that provably fair. What do we use to pick the winner? I guess the same issue exists for the "roll 99.9999 to win" jackpot idea - I could brute force myself a client seed that rolls a 99.9999 whenever I wanted if I wanted to claim the jackpot for myself. Maybe jackpot/lottery schemes can't really be provably fair. The way I've seen it done before it to use an external source of entropy: "the hash of the next bitcoin block found after midnight" or whatever.

But even then... so we decide that we're going to use the last 8 hex digits of the hash of the next bitcoin block to tell us which bet wins the jackpot. We can divide those last 8 digits by 0x100000000 to get a random number between 0 and 1, and use that to pick the winning bet. Suppose there were just 4 bets, worth 1, 2, 3 and 4 CLAMs, and the BTC block hash gives us 0.64653 for the 0-1 ranged number. We'd line the bets up in betid order (4, 1, 3, 2), sum them (10 CLAMs), multiply that sum by the BTC block hash number (0.64653 * 10 = 6.4653), then step through the bets in order to see which one takes the total to that amount. The 2nd bet takes the total to 5, and the 3rd bet takes it to 8, so the 3rd bet wins. That seems like a fair way of doing it, but is impossible to verify without a list of all the bets. And there are 5 million bets per day, so that's not practical. I guess if we arrange the bets by userid and sum them by userid it becomes much easier: (eg. the 1 and 2 were by user 23, and the 3 and 4 were by user 21. We sum by userid and list them in order of userid, so the list becomes just: {user21: 7, user23: 3}. Then user21 wins, because he takes the total bet over the 6.4653 threshold. That means that low betids win more when the last 8 digits of the BTC block hash are low, which is potentially gameable by large miners, but it's probably not worth worrying about for the size of jackpot we'll be talking about.


We can pay out small amounts regularly, like 1% per hour maybe, and 5% per day - then have bigger weekly and/or monthly payouts.

Thanks for your feedback. I'm told that Scabby had some interesting things to say about a possible jackpot scheme in the JD chat this morning, but I missed it. I'll have to go back and read what he had to say.

I deleted most of both of our posts and left the bit you seem to have missed. I'm proposing that I fund the jackpot.


If you hit the target roll number when betting 100 CLAMs or more, you win the whole pot, yes. I figure that would be very rare. People don't often bet 100 or more CLAMs, and hitting it would be a million to one shot. I would fund the pot myself, and I guess allow people to tip the pot, like they do for the bet milestone prizes.

Good point. The site sees something like 20k CLAMs wagered per day on a slow day. The house edge of 1% means we expect to win 200 CLAMs of that, and the 10% commission the site takes means that the site expects to keep 20 CLAMs from that. That's not enough to excite anyone is it, even if we put it all in the jackpot pool.


Thanks.

We hit 100k profit for a few seconds just now for the first time:

So a couple of days ago, DaDice added a jackpot feature to their site. It seems particularly poorly thought out, but it got me thinking...

How *should* a jackpot feature be implemented?

The DaDice jackpot has several flaws:

1) it's not provably fair at all - it depends on both the betid and the time, both of which can be undetectably maniputulated by the site

2) their jackpot has the best edge for the smallest bets; it actively encourages people to make tiny dust bets rather than big bets

3) their jackpot has a huge negative house edge; it is very gameable and unsustainable, unless they intend to cheat (see point 1)

So how can we do it better?

How about:

1) Since we have 6 digit rolls as opposed to their 4 digit rolls, we can get rid of the betid and current time components, and just make it that you need to roll 99.9999 to hit the jackpot. That makes the jackpot a 1-in-a-million chance. We see around 5 million bets per day, and so will see that many jackpot wins per day.

2) Clearly we need to scale the prize according to the amount risked. It's silly to encourage dust bets. And so hitting the jackpot will pay out a share of the prize pool based on the site of the bet. How about if the bet size in CLAMs is the percentage of the prize pool you win. So if you hit the jackpot with a bet of 5 CLAMs, you get 5% of the pool. That way the pool doesn't shrink too fast and gets a chance to build up. The vast majority of bets are dust, and so we can expect most jackpot wins to be bets staking less than 1 CLAM, and so it will be relatively rare that the pot goes down by much.

3) We've promised investors a 1% house edge, and 10% commission. The house will have to fund the jackpot itself since we don't want to change the deal we made with investors. We could initially fund it with 1000 CLAMs, and leave it to stake. Would the staking rewards make up for the payouts it makes? Staking pays out 0.2% per day, and we expect 5 jackpot wins per day. So long as the sum of the stakes of the 5 winning bets is less than 0.2 CLAMs, the pot will increase that day. That's probably usually going to be the case.

Thoughts? Suggestions? How would a well thought out jackpot system work? Is the proposed scheme worth implementing? Are the proposed prizes too small to bother with?

I can try to explain how it works...

Replace BTC with LTC or DOGE throughout, because it applies to them just the same... for the sake of simplicity I'll assume you're importing a BTC private key.

When the initial distribution was made, the CLAM developers looked at the BTC blockchain, and made a list of all the addresses that had funds. They converted each BTC address into the corresponding CLAM address. The only difference is the "version number" that is used to convert the RIPE-MD hash into an address - in BTC it's version 1 (which makes BTC addresses begin with a '1') and in CLAM it's version 137 (which makes CLAM addresses begin with an 'x'). Then they funded each of the corresponding CLAM addresses with 4.6 CLAMs.

To spend those 4.6 CLAM outputs, you need the addresses' private keys. The private key for each CLAM address is the same as the corresponding BTC address' private key. To spend an output, you create a transaction that moves the value to a new address, and sign it using the private key. Then you broadcast the signed transaction to the network. The private key itself is only used for signing and isn't broadcast.

Whether "the keys are secure" or not depends on whether the CLAM client software is trustworthy or not. You are giving it your private keys, and trusting it to only use them to sign a transaction that moves your 4.6 CLAMs for you. In theory it is possible that it could also check whether there is a balance at the corresponding BTC address. It could, in theory steal any such BTC. There have been no reports ever of the CLAM client software stealing anyone's BTC, so it's likely safe. Also the source code is open source. You can inspect it, and build the client for yourself from source. That's what I do.

Incidentally, note that any random shitcoin wallet you download is just as able to steal your BTC. When you download and run a program, it can do whatever it wants on your computer. There's nothing to stop it looking in your bitcoin folder and uploading your BTC private keys to its author. There's nothing to stop it watching you type your wallet passphrase into the Bitcoin wallet, and then telling the passphrase to its owner. The message here is to be careful *whenever* you download and run programs from the Internet. Any program is potentially coin-stealing malware.

This probably isn't what you were hoping to hear, but it's the truth. Downloading and running programs basically allows the program's author to do whatever he likes on your computer, whether or not that program has an "import wallet' menu entry.

Personally I trust the CLAM client, and have imported wallets which have hundreds of BTC still in them without any issue.

True. But it's better still to play a single martingale sequence than to make a single bet.

Instead of betting 2 units at 1.5x (66%) to get to 3 units, with a chance of success of 66%, try betting root(3)-1 units with a multiplier of (3 + root(3))/2, and if you lose, bet the rest at the same multiplier. If either bet wins, you'll have 3 units. Your chance of success is 66.1768%, which is higher.

The single martingale sequence has a lower expected risk, and so has a lower expected loss.

I think the unofficial page looks better and is easier to navigate too.

I recently found this CLAM faucet: https://cryptospout.com/faucet/hourly_clamcoin_faucet.html

It has a bunch of CLAM links, but I see the "Website" link goes to one of the unofficial CLAM pages. There's a link to submit proposed changes to the links, so maybe it would be worth fixing and/or adding links.

So do I need to withdraw and redeposit after every time I use the faucet to be eligible for the jackpot?

And if I withdraw and redeposit the same amount 10 times, so all 10 deposits count to the total deposited?

I'm wondering what you are trying to encourage with these odd rules.

Well that addresses all my concerns then. Thanks.

It would be better if you could find a way of making the jackpot provably fair.

Under the current rules it would be trivial for you to delay a bet by a fraction of a second to ensure its betid is non-prime if the "wrong" player is about to win it.

Why have this prime number requirement at all?

edit:


Do you realise that this means the house edge is smaller for tiny bets than for big bets? It will encourage smart players to make smaller bets.

Finally, since the Unix time is prime about 1/22 of the time, and the betid is prime about 1/20 of the time, and you hit the required number 1/10000 of the time, betting 1 satoshi will hit the jackpot about 1 in 22*20*10000 bets. That's 1 in 4.4 million bets. But the jackpot is 20% of 1 BTC, or 20 million satoshis. That makes the jackpot very +EV:

>>> 100 * (1 - 20/4.4)
-354.5454545454545

The house edge when betting 1 satoshi is minus 354.5%...

I'm not confused. I made some guesses about what the exploit may have been. There's no way to know for sure without inside knowledge of the Prime Dice code.

Don't forget to move the git tag if you didn't already. I seem to remember you had to make a new commit to get it to build on Windows but didn't move the tag.

Did this happen yet? It has been 10 days since you wrote that.

The message suggests it's unable to open the leveldb. I can't imagine that stuff changed between v1.4.7 and v1.4.10, so I'm stuck for an explanation.

I don't have the same problem, but I'm not using Windows. I build my own executables.

Can anyone confirm that the 64 bit v1.4.11 Windows executable linked in OP works for them, and whether it works on a blockchain written by v1.4.7 previously?