Do *lucky* plots exist?

The documentation for the check command seems to hint at differences in plot quality.

  • Is this true? Do some plots contain more potential quality proofs than others? I don’t mean just returning a better ratio when tested through check, but actually a difference in the total amount or the quality of the potential proofs contained.

  • If so, could someone explain why? what is the underlying basis? Are the proofs less orthogonal to each other? Too many near duplicates etc?

  • Suppose you have plotted all your disk space, would you consider removing and re-plotting the worst plots with the idea of gradually over time upping the average quality of your farm?

I have always wondered about this too.

Even at -n 300 10% differences exist, so that is not nothing if this would be relevant

No, it’s myth. Check this:

I know that video, but I’m not sure it answers my question.

  • If the ratio is <1, your plot was relatively unlucky.
    • This shouldn’t really concern you unless your ratio is <0.70 # If so, do a more thorough chia plots check by increasing your -n

Page 27 of the PoSpace PDF says the answer to “why” is random noise. In other words, Number Theory.


This sample size calculator shows your n=300 test is adequate:

I just tested a plot,
for n=30: 1.1
for n=666: 0.991
So it takes many tests to reach high confidence intervals.


The test challenges are not automatically randomized between test runs (if you run the check again for the same n you’ll get exactly the same result).

So if you want to use different challenges in your next test, use an offset greater than or equal to your n:

“chia check plots -n 30 —challenge-start 30”

You’ll want to use the -g plotname option to avoid checking hundreds or thousands of plots if you have that many.

I believe some plots can be corrupted, I’ve never heard of unlucky plots.

At current costs I wouldn’t replot specifically for that, but if I had to replot some drives I would try to diagnose plots to make sure I didn’t have any bad luck plots (if there’s such a thing) : run about 4 or 5 tests on different parts of the plot and for 300 proofs.

I always make sure that all my plots are good after plotting. If I had a full ram plotter, maybe the electricity cost of replotting would be worth it, in the long run on a big farm where the plotters are sitting idle.

1 Like

Here I found a plot that is 2.6% more lucky than expected. It’s my largest k=32 plot.

You really believe that plot check of 6000 is really a significant indicator for a plot file that contains at least several million potential proofs? Also having higher than the number of checks means that it found multiple proofs for the same challenges at the set difficulty, this could actually be considered a negative because your plot is less “diversified” and could be seen to contain fewer unique proofs. The entire idea of plot quality being measurable is ridiculous.

I’m going to run the test even longer for more accuracy. No it’s not ridiculous, it’s mentioned in the official technical documentation.

Take that with a grain of salt. Also, it looks like you’re running chia plots check, which is slow as dirt because it uses python. If you’re serious about testing plots you should download and build the real chiapos repo which is written in C and is significantly faster.

The python code calls the chiapos c++ code.

You’ll have to test yourself, but in my testing running chiapos directly was significantly faster.

As the link I posted to the PDF above states, the random luck difference is small, but it does exist.

The subject of this thread isn’t about you attacking the Chia Inc implementation of plot checking.

Ok dude, you believe whatever you want to believe and waste your time and resources chasing this phantom, I’m done trying to help you.

As far as interesting improvements are concerned, collecting statistics continuously while harvesting would be a good optional feature.

1 Like

You are now muted since you are only wasting my time. I won’t see your replies. @gryan315

1 Like

This is why the answer to the question of “do lucky plots exist?” is NO. When you are checking plots, it is just using a “hard-coded” challenge. So your plots might be “lucky” or “unlucky” for that specific challenge but we don’t know if they will be lucky or unlucky for the real challenges issued by the blockchain. If they are lucky for a specific challenge, great, you win the block! If they are unlucky for that challenge, you don’t win the block.

Checking plots gives you the confidence that your plots will probably also pass real challenges, but it gives you absolutely no indication of if that plot will be “lucky” for any other challenge. Therefore, the question “do lucky plots exist” doesn’t really make sense - of course they do - they are the ones that win the blocks! But also once the block is won, any other plot might be the new “lucky” one.

1 Like

Luck as measured in the reference, is the number of proofs per challenge.

Block challenges are random and uniformly distributed.

You are redefining luck to relate to specific challenges.

Yes, I think we are saying the same thing, but differently. The number of proofs per challenge will obviously depend on the challenge. When you check plots with the CLI, the challenge is always the same (unless you use the offset, which I didn’t know about - cool!). Therefore, it wouldn’t make sense to determine “luckiness” from running plot checks. A plot may have a lot of proofs for the test challenge, but none for the “real challenges” over the next month. Does that make sense?