[CLOSED] BTC Guild - Pays TxFees+NMC, Stratum, VarDiff, Private Servers - page 232.

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Quote from: eleuthria on September 28, 2013, 02:36:35 AM

Quote from: organofcorti on September 28, 2013, 02:32:05 AM

Disregard most of the last post - I have another look at your chart and it already is labelled with blocks solved. Sorry about that.

Anyway, having the x axis scaled to blocks solved and the y-axis log scaled means that luck will be represented without bias in either direction. Nice job, mate.

X-axis is actually not scaled to blocks solved. Y-axis is the luck (% of payout vs expectation), X-axis is shift ID (roughly time since shifts are normally the same length).

OK, you could keep it as is but scale the distance between the tick-marks to the number of blocks solved within a shift - that would make the luck graph as accurate as possible and eliminate the influence of the increased round durations during times of bad luck. It might make the charts less intuitively understandable though, and I'm not sure how easy it is to create a custom scale transform in js.

eleuthria

legendary

Activity: 1750

Merit: 1007

Quote from: organofcorti on September 28, 2013, 02:32:05 AM

Disregard most of the last post - I have another look at your chart and it already is labelled with blocks solved. Sorry about that.

Anyway, having the x axis scaled to blocks solved and the y-axis log scaled means that luck will be represented without bias in either direction. Nice job, mate.

X-axis is actually not scaled to blocks solved. Y-axis is the luck (% of payout vs expectation), X-axis is shift ID (roughly time since shifts are normally the same length).

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Quote from: eleuthria on September 28, 2013, 02:02:57 AM

Quote from: organofcorti on September 28, 2013, 01:55:58 AM

Quote from: eleuthria on September 27, 2013, 12:41:33 AM

One thing I think the chart will highlight is how PPLNS smooths variance. You won't see big spikes/cliffs of luck that only last a short burst and go way positive/way negative. Instead you'll see it trend positive for a while, then negative, since PPLNS is designed to basically even out per-round variance by stretching it over the last N shares (in this case, broken up into shifts).

That's a great chart - the only comment I have is that you should probably have it on a log scale rather than linear - on your scale luck can be infinitely high, but can not reach 0. Using a linear scale gives more weight to the positive luck.

Bad luck is already going to be disproportionately represented on a graph due to bad luck being defined as "long time without blocks", so either way it's going to be biased.

Well, that's another problem with averaging luck over time instead of blocks. Over a longish period of time there's no significant effect, but over a day averaged luck will be, on average, a few percent lower than the real luck. The only way around that problem is to average over "n" blocks and but label it with dates. The only problem is changing the number of blocks over which the luck is averaged if the pool loses or gains large percentages of the network.

Edit:

Quote from: eleuthria on September 28, 2013, 02:18:11 AM

Okay, after playing with it, I went with a log scale on the graph

.

Disregard most of the last part - I have another look at your chart and it already is labelled with blocks solved. Sorry about that.

Anyway, having the x axis scaled to blocks solved and the y-axis log scaled means that luck will be represented without bias in either direction. Nice job, mate.

eleuthria

legendary

Activity: 1750

Merit: 1007

Okay, after playing with it, I went with a log scale on the graph

.

eleuthria

legendary

Activity: 1750

Merit: 1007

Quote from: organofcorti on September 28, 2013, 01:55:58 AM

Quote from: eleuthria on September 27, 2013, 12:41:33 AM

One thing I think the chart will highlight is how PPLNS smooths variance. You won't see big spikes/cliffs of luck that only last a short burst and go way positive/way negative. Instead you'll see it trend positive for a while, then negative, since PPLNS is designed to basically even out per-round variance by stretching it over the last N shares (in this case, broken up into shifts).

That's a great chart - the only comment I have is that you should probably have it on a log scale rather than linear - on your scale luck can be infinitely high, but can not reach 0. Using a linear scale gives more weight to the positive luck.

Bad luck is already going to be disproportionately represented on a graph due to bad luck being defined as "long time without blocks", so either way it's going to be biased. Given BTC Guild's settings (N ~= 2,500,000,000), I think linear will work out fairly well. The real important part (IMO) is the two lines cutting across showing the average vs neutral.

I will look at some alternate scaling methods, but like I said, with such a high N value, there is an "effective" ceiling on how high luck can possibly go (outside of struck by lightning 19 times while holding a winning lottery ticket types of luck).

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Quote from: eleuthria on September 27, 2013, 12:41:33 AM

One thing I think the chart will highlight is how PPLNS smooths variance. You won't see big spikes/cliffs of luck that only last a short burst and go way positive/way negative. Instead you'll see it trend positive for a while, then negative, since PPLNS is designed to basically even out per-round variance by stretching it over the last N shares (in this case, broken up into shifts).

That's a great chart - the only comment I have is that you should probably have it on a log scale rather than linear - on your scale luck can be infinitely high, but can not reach 0. Using a linear scale gives more weight to the positive luck.

-Redacted-

hero member

Activity: 574

Merit: 501

Nice chart - good change. Presents luck well.

demonmaestro

hero member

Activity: 574

Merit: 500

Mining for the hell of it.

I like the overall average. and I really like the luck having as of currently.

However with my wopping 12GH/s i aint going nowhere fast.. lol Cheesy

Come on KNC.

Holy snake batman. 149% and climing!

eleuthria

legendary

Activity: 1750

Merit: 1007

One thing I think the chart will highlight is how PPLNS smooths variance. You won't see big spikes/cliffs of luck that only last a short burst and go way positive/way negative. Instead you'll see it trend positive for a while, then negative, since PPLNS is designed to basically even out per-round variance by stretching it over the last N shares (in this case, broken up into shifts).

demonmaestro

hero member

Activity: 574

Merit: 500

Mining for the hell of it.

Quote from: eleuthria on September 27, 2013, 12:26:29 AM

I'm definitely looking at updates/revisions to the charts. Only issue with a "total speed" line is it won't be on the same y-axis scale, otherwise accounts with many workers will just have all their workers flattened at the bottom due to having to fit a total line in there. Makes things a little confusing IMO.

Good point. I like the graph for the PPLNS! Shocked

BTW i see 130.61% Grin

neutralist

member

Activity: 85

Merit: 10

Quote from: mdopro1 on September 26, 2013, 10:46:32 AM

Pool luck 62% ?? I've never seen it this low with BTCGuild. Did difficulty change or something?

60%? I have seen 38%. may be you want to try out PPS.

eleuthria

legendary

Activity: 1750

Merit: 1007

Quote from: demonmaestro on September 26, 2013, 11:35:43 PM

Thanks..

I guess this is wrong.. lol

Quote from: demonmaestro on September 26, 2013, 02:13:02 AM

Thus confirms my speculation that eleuthria never sleeps. Cheesy

No fret!

Eleuthria: Just an idea but could you possibility have a total hashing power line on the user charts?

I'm definitely looking at updates/revisions to the charts. Only issue with a "total speed" line is it won't be on the same y-axis scale, otherwise accounts with many workers will just have all their workers flattened at the bottom due to having to fit a total line in there. Makes things a little confusing IMO.

demonmaestro

hero member

Activity: 574

Merit: 500

Mining for the hell of it.

Thanks..

I guess this is wrong.. lol

Quote from: demonmaestro on September 26, 2013, 02:13:02 AM

Thus confirms my speculation that eleuthria never sleeps. Cheesy

No fret!

Eleuthria: Just an idea but could you possibility have a total hashing power line on the user charts?

jarhed

sr. member

Activity: 672

Merit: 254

Quote from: demonmaestro on September 26, 2013, 09:19:05 PM

You know you're a server admin when...

FTFY

....never sleeps Huh

demonmaestro

hero member

Activity: 574

Merit: 500

Mining for the hell of it.

Quote from: eleuthria on September 26, 2013, 08:02:19 PM

I'm going to be playing with the charts even more this weekend. Now that we're on the new DB server, I can start playing with a lot more data without worrying about the poor thing having a seizure.

You know you're a server admin when... Cheesy

eleuthria

legendary

Activity: 1750

Merit: 1007

Quote from: OleOle on September 26, 2013, 07:56:44 PM

"A picture tells a story of 1000 words."

Thanks for the new graph at the bottom of the PPLNS Stats page, it makes it much easier to see how things have been tracking with luck.

Wink

I'm going to be playing with the charts even more this weekend. Now that we're on the new DB server, I can start playing with a lot more data without worrying about the poor thing having a seizure.

OleOle

hero member

Activity: 532

Merit: 500

Are you like these guys?

"A picture tells a story of 1000 words."

Thanks for the new graph at the bottom of the PPLNS Stats page, it makes it much easier to see how things have been tracking with luck.

Wink

mdopro1

sr. member

Activity: 308

Merit: 250

Double your Personal Bitcoin Funds.

Quote from: eleuthria on September 26, 2013, 11:15:20 AM

Quote from: mdopro1 on September 26, 2013, 10:46:32 AM

Pool luck 62% ?? I've never seen it this low with BTCGuild. Did difficulty change or something?

This isn't unusual. Luck is now shown on a per-shift basis in the chart, rather than a per 10-shift basis. Even so, we've had perios where TEN shifts averaged out to under 60% luck before. We've also had luck at 180% for 10-shifts. You'll notice our average is still quite a bit above neutral at the moment.

Still pretty good and I see it is improving too. I want to see 180% for 10 shifts again... can you press the easy button?

Quote from: dlasher on September 26, 2013, 11:25:31 AM

Yes, diff just jumped from 112M to 148M.

Code:

Blocks 260250
Total BTC 11.756M

Difficulty 148819200
Estimated 152406811 in 1830 blks

Network total 1225.272 Thash/s
Blocks/hour 6.90 / 522 s

Means I need to get more ASICs.

dlasher

sr. member

Activity: 467

Merit: 250

Quote from: mdopro1 on September 26, 2013, 10:46:32 AM

Pool luck 62% ?? I've never seen it this low with BTCGuild. Did difficulty change or something?

Yes, diff just jumped from 112M to 148M.

Code:

Blocks 260250
Total BTC 11.756M

Difficulty 148819200
Estimated 152406811 in 1830 blks

Network total 1225.272 Thash/s
Blocks/hour 6.90 / 522 s

eleuthria

legendary

Activity: 1750

Merit: 1007

Quote from: mdopro1 on September 26, 2013, 10:46:32 AM

Pool luck 62% ?? I've never seen it this low with BTCGuild. Did difficulty change or something?

This isn't unusual. Luck is now shown on a per-shift basis in the chart, rather than a per 10-shift basis. Even so, we've had perios where TEN shifts averaged out to under 60% luck before. We've also had luck at 180% for 10-shifts. You'll notice our average is still quite a bit above neutral at the moment.

Topic: [CLOSED] BTC Guild - Pays TxFees+NMC, Stratum, VarDiff, Private Servers - page 232. (Read 903150 times)