Miners - I'd like to know your opinion


#1

Good morning everyone,

At this juncture I’ve been watching Neos mining trends for some time in effort to make a determination as to what might prove more beneficial for the future of Neos and longevity. I’ve noticed there are never any miners on X11 just barely, and SHA hash has dwindled down to nearly nothing. What’s your take on the cause/reason for this? You guys are out in the trenches mining, we’re here doing what we do - so we’d love to hear it from the people that it matters most to - the miners. Everyone of course is welcomed to speak their piece regardless of miner status or not of course.

Best regards,

syntaks


#2

Doesn’t the difficulty of each algorithm keep the ratio of miners balanced? If nobody is mining X11, couldn’t one guy with a large rig jump in and make a killing with a low diff?


#3

Hmm “balanced” to a point, but if the diff is driven high, and then hash pulled off, it’s stranded. This is the reason we’re also looking into new diff algos to prevent that. I’m wondering why noone’s mining on sha/x11 myself honestly when the opportunity is there.

Best regards,

syntaks


#4

i always found the retargetting to be weird …

wouldnt it be logical for when a block is found on algo 1 that diff on algo 2 and 3 should drop to something lower? or am i looking at it weirdly …


#5

I always regarded mining to be a kind thing of secondary importance for one coin. I think energy and resources should be redirected to things that are of a more benefit for consumers and merchants, not miners. Saying that, I just wish to state there’s one area in which there is still low competition, that is instant/faster transactions.

Me I’m mining with 200 GH/s on mp.neoscoin.com and I’m kind of lazy to even think about how much coins they made - Avalon1 is already technologically dead.

Guys, I lost like 9k USD on Neoscoin. It’s only my fault, but this project really looked like promising one and I still have faith. Lost even more on Vertcoin, the difference is VTC is absolutely hopeless.


#6

Good Afternoon,

Miner here (30 GPUS, have mined ~40k Neos since launch …sold 20k currently holding 20k)
I also mine other coins I run python scripts that switch different coins based on profitability. (Feather, DGB, Myriad or whatever shit-coin is launching that I can make a quick buck off of it, shameless but sorry it’s fun to play around with for me)

I default to mining NEOS because your coin is very impressive compared to the field; it’s one of only 5 coins I hold. Ok so with that out of the way let’s get inside the head of mining with regard to x11.
I technically have a line of code that would mine via x11 but the difficulty relative to the blake256 would need to be less than about 100 for it to be more profitable, as such my rigs keep mining Blake not x11. I also don’t have any FPGAs that grant advantage in x11 leaving with the Blake algo as my only option. You can check verify my behavior as both of us mined at the same x11 pool but no longer do, my handle (Bagdar13 and I am assuming your handle of syntax are the only two that found blocks for a long time).

So, in terms of multi algo implementations as a miner, I am highly impressed with DGB they seem to have stabilized a bit and get 20% each algo per day (mind you they used 5 algos and I believe 3 of them are GPU only, I am guessing this is why you want to change the number of algos implemented and which ones are in use). I think this matters considering SHA256 can be dominated by 1-2 high end devices, also there is a lot of x11 . In the first few weeks of mining it was profitable to just spend a few btc on nice hash run x11 through a pool and crack a quick 10 blocks in under a minute.

Ok, so the reality is most miners don’t know how retargeting works, I have even found that some devs don’t understand why retargets overshoot and undershoot., It seems to me that you have a pretty good understanding of what you want to happen but didn’t actually model the behavior of your algorithms when hit with an exogenous shock (e.g. miners coming online and offline). I believe your implementation works well under steady state. However, consider that pools and miners like myself come online and offline based on the difficulty

This is a classical controls theory problem. It is also easily simulated using System Dynamics; checkout the program vensim if you want to construct simple models.

Without getting too much into the details effectively the length of the memory in relation to the rate at which blocks are being found (and I believe you capped the change at a max of 25% in a stability vs performance trade). In steady state (if you imagine the algo difficulty to look like a wave) if all three algos had a constant amount of hash sent at them they would indeed stabilize at a near constant level difficulty. (or at least would be in small waves of difficulty)

The response of a system to shock (imagine a spring) can either be dampening, oscillatory, or expanding/growing. Remember as I earlier said the miners coming online and offline with regard to the difficulty are the exogenous shock to the system. The miners disappear however, due to the number of blocks in the memory for the difficulty adjust forces the algorithm to “crest” up to the top of the wave but then the miners go offline. Depending on how this behavior aligns across the three algorithms this can produce a contributory effect in both the increase and decrease of the difficulty across the algorithms forcing to unreasonably high or unreasonably low difficulties.

There is a much deeper discussion here, let me know if you want to pick apart or dive deeper on the problem. As you already know it is not as simple as shortening the number (or removing the 25% change cap) and the reason for this is you then trade stability back for performance and can create an even more unstable system. This problem is no different than the control system for an airplane being buffeted in the wind.

One modification that is easy and can help with dampening is similar to the “uptick rule” that was implemented on the stock marked back during the crash. Effectively, if the difficulty is going down or up, implement a rule that says every X number of blocks cannot be in the same direction, so if say four block decrease in difficulty the fifth block must increase in difficulty before another decrease is allowed. This adds stability without sacrificing too much performance. What you don’t want is to let people like me ride the coin down and then force it up. If you implement this on both sides across all algos it should provide a partial solution.

Simulate that and let me know what you think

Also, the reality of large hash-power as an exogenous shock must be considered. Likely it is impossible to code something that can protect against terahash level power in sha or x11. This leaves with a philosophical debate, go to gpu/cpu sytle only algos or play in the big leagues knowing the coin will be quick mined and dumped by farms (and I don’t mean small timers like myself with 30 gps). There is clearly a desire on both ends of the spectrum. (I think DGB with 5 algos actually shielded against this but you can see this risk manifest in digihash.co where one sha256 miner gets 20% of the entire network reward and it seems like that guy is dumping). Sorry I don’t have any idea what the right answer is here, but we can theory craft more if you like.

Have a great day and I must say I am really impressed with you as a dev!


#7

Man, I can’t begin to express my sympathy and apologies for your loss or my understanding first-hand about it being as I bought Neos myself around 35-36k and never sold a single one and I still won’t and wouldn’t. What I can promise you isn’t that we’re going to make empty promises to rally up people into buying into hype - but that we’re going to continue to be here every day and on top of current issues and trying to improve every facet of Neos possible. I’m about to respond to the gentleman who responded after you about this - but right now the diff algo is in our crosshairs in terms of figuring out a better approach (which he has mentioned). We’re not stopping - period. I thought at first it was a botnet mindlessly dumping. I’m in my right mind to think however it’s a mix of poor distribution via mining and intentional price suppression. With 2.0.2 in the works, a couple of additional projects that are currently in beta - I strongly feel we’re only in the direction of vast improvement. I’ve reached out to PR firms, and priced out advertising options through certain media. The new approach in 2.0.2 is worth us firing on all cylinders then rather than putting in a huge attempt now in my opinion. If I’m wrong and the community tells me I am that means one thing - I have to work harder. We’re looking into alternatives now since our approach last week didn’t really prove to be worthwhile. This time around though we have new options we’re going to be testing with. If those pass tests, we’ll make everyone aware of the changes and any possible extra information they might want to know. In the meantime, the only thing I can say to you - is thank you, for your faith and for your continued support. Like I’ve said many times over, the community makes Neos not the other way around.

Sincerely,

syntaks


#8

Bagdar,

First let me say that my shortening of your response is in no means an attempt to show any disrespect whatsoever, I just didn’t want to make my response on top of yours create an incredibly long post. Secondly, in reading what you’ve written I can say i wholeheartedly agree with you. I also respect your vast knowledge in this area and appreciate you sharing it with the rest of us. Thank you very much. The breakdown of why you’re mining blake and not x11 gives us some indication as to what others might also be thinking and some cause for the current mining statistics. We’re looking into replacing or adjusting the existing diff algo for better responsiveness to the exact situation you’ve described. We’re also considering replacing x11 with perhaps scrypt. I’d love to hear thoughts on that from folks as well. Thanks again for your compliments and support. We wouldn’t be here without you guys.

Sincerely,

syntaks


#9

I decided to switch my x11 away from the multi for now onto the main pool (ka pools) and it’s only a lowly 3mhs but I’m hoping we get a little lucky and nudge the difficulty into re-targeting. Is that how it works?


#10

No worries,

No worries no need to quote me :wink:

A few more thoughts…so unless we are radically going to shift how we recomputed the concept of difficult and block time, which I don’t think is what we are after, the issue of the diffusely readjust algorithm, is at its core, an “information delay.”

If we look at Bitcoin it is no more than a “first order” information delay. That is to say it just looks at average speed over a time period and recomputes the new difficulty (and it has always been going up). It is a simple yet efficient algorithm that works under the assumption that no massive swings will ever affect the network. While I haven’t done an inept look at the exact parameters it seems that Bitcoin would be robust in change to between 25-33% increases over the timespan of a week; again that’s just a wag, a change ± in hash power on the bitcoin network would open it to manipulation of difficulty as seen in altcoins, my point is it is not immune, it is the economics that makes this form of re-targeting possible right now not the code.

When we look at other readjust algorithms such as gravity well (and its flavors dark etc) as well as digishield etc what those are effectively are multiple information delays that are tuned with the expectation of manipulation by miners. See http://bitcoin.stackexchange.com/questions/21730/how-does-the-kimoto-gravity-well-regulate-difficulty for more information. Still this is just a smoothing function with different time parameters. I have seen asymmetric readjust algorithms but I don’t think anyone has solved the question of how do you bring a new coin to market with the existence of large computational power.

Your difficulty retarget algorithm is an information delay that you have tuned as best you can to mitigate the expected behavior of miners. However, it is still a simple information delay, that performs a buffering action.
After giving this some thought over the last few hours I have constructed an argument that it is not in the best interest of a coin like Neos to use SHA 256. By analogy take an airplane, no matter how good an airplane you build, you cannot create an autopilot that can counteract the strongest possible wind. In this analogy imagine an airplane hit by a tornado, that’s what the power of sha 256 is equivalent to these days hitting your information delay/readjust algorithm. If your readjust algo is known and the coin profitable to mine then miners can exploit the readjust algo and dump coins; we can make a very sturdy airplane(very robust code) but eventually the overwhelming hash power will be the dominant factor.

The argument as I understand it is for a coin to gain adoption we want to use the major algorithms in play sha-256 and x11 etc. However, we get insta-minned and dumped if we use those algos…its a losing proposition. In the case of DGB with 3 “arcane” gpu only algos they are still losing 20% of their coins per day to SHA256 which forms in my mind an argument for dropping sha-256.

This is a similar problem we are seeing with x-11 but because there is less power in x11 no one stay on x11 to ever bring back down the difficulty. Which has made x11 effectively nonfunctional, leading us into this situation of flash-mine sha256 about 2 hours a day and then a large oscillatory wave on the Blake difficulty.

Effectively we likely need to drop any algo where 1 person (or small group of people) could take over; as they will dump coin value. At the very least to grow we want people to mine. Interestingly enough, my argument actually implies that to grow a coin you should start with “rare” algorithms, and then add in mainstream algorithms after the coin has gained adoption. This also implies a roadmap of hard-forks, which is scary to some people as hard fork requires major community support (and fear). Effectively, adding on more mainstream algos as the coin reaches maturity and can handle the system shock. The coins that seem to be doing the best are those where people mine just below profitability but stay in for maturation.

I will sleep on this problem and see if I can construct a more concrete argument, for now if you were to make a best algo list here are my thoughts.

Neos does not need to be in the business of algorithm improvement/generation etc but we must respect the role algos have in the coin, also picking algos that are actively being developed is probably a good idea too, no one is working on better saffron Blake miners.

Drop x11, SHA-256 and Blake. (if you have to keep one keep blake but it is a “hot algo” which pushes miners away and is not actively developed)

Two thoughts:

  1. A CPU only algo (there are several to choose from, also this was your plan at launch but there was an existing saffron Blake miner at launch)
  2. Neoscrypt; tying to this might have some positive publicity.
    Bottom-line if I can buy hash for .05 BTC on nice hash rout it directly to hash harder and then dump those same coins on bittrex for .2 BTC the economy will be hurt.

I realize what I have just written effectively messes up all the pools in existence etc (but x11 is broken already). Where we are at right now, probably requires getting our ducks in a row figuring out exactly the role of mining in relation to NEOs and then doing a relaunch party with a long lead time. Likely this need to be pondered as well.

Sorry I didn’t see what algos you are considering and the logic behind each, is there another thread?

EDIT: this is respect to the core coin only not the multipool; its great to have as many algos as possible on the multipool, and puts up-pressure on the coin price.

Thoughts?

Have a good day,


#11

If you want to see decision logic here it is: (edit this is based on a 7870s speeds)

coin1_hash = 1000
coin2_hash = 3.4
coin3_hash = 165
coin4_hash = 14
coin5_hash = .15

coin1_prof = 2000
coin2_prof = 20
coin3_prof = 20
coin4_prof = 20
coin5_prof = 8000

coin1_time = 1
coin2_time = 1
coin3_time = 1
coin4_time = 1
coin5_time = 1

coin1_reward = 50
coin2_reward = 7300
coin3_reward = 7300
coin4_reward = 7300
coin5_reward = 80

coin1_f = ((coin1_rewardcoin1_prof)/coin1_time) / (network_diff/coin1_hash)
coin2_f = ((coin2_reward
coin2_prof)/coin2_time) / ((network_diff2/256)/coin2_hash)
coin3_f = ((coin3_rewardcoin3_prof)/coin3_time) / (network_diff3/coin3_hash)
coin4_f = ((coin4_reward
coin4_prof)/coin4_time) / (network_diff4/coin4_hash)
coin5_f = ((coin5_reward*coin5_prof)/coin5_time) / ((network_diff5/64000)/coin5_hash)

coin 1 is neos 2-4 are DGB, 5 is feather, i mine whoever has the highest f factor and I do it on a prop pool so that i get counted only in the current round.

coin_time is one because block time is always broken with multi algo coins, and right now people are manipulating feather so its going form 25-75 in a 2-3 hour cycle.(i point this out as single algo does not protect against manipulation).

This stuff is really cool,


#12

6paqneos you are a trooper! Sadly, that wont help because even if you did get it down to parity with sha 256 then a large pool would come online and the difficulty would be back above 500 in under 10 minutes.

Also, your time to find a block is 2^32*900 /3000000 m/hash /60 seconds /60 minutes or 357 hours…


#13

Well shit…I switched back! hahaha. Thanks for the heads up. Makes more sense to stay on the mp.


#14

Thanks a lot for the feedback folks. Someone I’ll make mention of that phark pointed out is that neoscrypt is gpu mineable now also. So that basically would remove the attraction of not enabling farms or certain groups to seek and destroy per se. At this point though, phark’s putting together a testbed with a new way to calculate the diff. In mitigating the massive diff swings and calculations we might come out ahead with a better result. Only testing will tell and we’ll let everyone know the results of it. Another thing that can’t be prevented, are these crazy spreads on the market and the way these trades are being executed in terms of people selling 500+ sats below top ask and selling into their own low bids.


#15

Well since the fork I can confirm that the difficultly has been between 1.5 and 2.3 million and as low 880 thousand. I have found many block at 750Gh/s looking good.