your logic points in 2 directions. i never claim improvement of luck. did i hit seven because i change pools? did i hit seven because i am using 99% of the space in all my hdd? or i am that lucky? i prefer to think is combination of all 3 and using a very efficient server now. In my 2nd video i show upgrades of all types of parts in my server that were slow or low performance. i have hit 2 more plots since then. i am not that lucky. i may go a year without hitting. regarless is just my opinion and you are free to think what you want. As far as data time will provide improvements in hitting blocks as i would see it in better than average block hits. you are not wrong…though
% was getting rid of plots that when checked shows as poor and added more usable spece in all my hdd. i never claimed it would make you hit more but rather it seems to improve my overall performance from a hardware view. no debates just my facts.
I was not debating anything. Rather I was asking if you knew how my farmr stats compare to your plot check results.
Of course, there are better and worst performing pools. Although, whatever pool you belong to, for me the first thing to do is to check your stale points. Those are due to either the farm not performing well (poor lookup times), and cleanup may help, or due to pool lackluster performance. In my opinion, those two are the biggest reasons to have “bad” luck. By switching pool, you potentially ended up in a better performing one, where more of your points are properly recognized.
My understanding is that you just slightly increased your farming size, so that was rather not a factor.
As far as replotting because of “bad” results from plot checks, that was already debated before, and I doubt it has any effects, just lost electricity. Although, it may be that some of the early plots may have issues, but I don’t know of any way to test for that.
So, that leaves us with mostly luck.
However, saying that, about a week ago, I also hit three blocks in a row (never happened before, never hit two in a row). Maybe during that time, the network was more congested, and a bunch of low performing nodes were hosed.
why does plot checks -n 30 has more sense in comparsion to plot checks -n 100 or even plot checks -n 500 ???
upd
i believe there is poor algorithm of checking itself, it has no random pick but some dichotomy alike routine, so with growth of N result will asymptotically seek to 1.
so “bad” plots is nothing but a superstition 
I dont really believe in this Plot quality thing.
But you can definitely have bad plots which have an error, cannot be read or what not. I for instance have 3 invalid plots which cannot be properly read. there is some error in the file data. I haven’t come to clean them up yet. Not worth it to replot them while im plotting anyways :D.
To recognize bad files, plot checks -n 5 should be enough already.
yes of course there are invalid plots or broken files/hdd but that’s not the topic of conversation 
that i agree with and to make sure plots are seen.
it is the one i choose to do initial look ups. i do also 100. 500 does not make sense to me because it may show plots as “good” but it could be a low quality anyway. it does not hust to improve the quality of plots.
Actually discovered 485 plots under .70 and 1 og plot that i didn’t even know I had. I an using now 2.59 tb of chia plots that was a waste of space before. Actually farmr found the OG plot for me.
all your points are good and valid. My experiment was not a waste of electricity or time. it was a logical step after my conversation with a chia expert who works at a chia massive warehouse in china. I import a lot from China and a product source , introduce his friend to me. We have been emailing each other, his english is hard to follow, but it’s been entertaining. As stated before never said it would guarantee blocks. You are right the larger your farm the better the odds. As far as testing, In my opinion, time will tell. I started farming in Aug 2021. I will wait until August of 2022 and compare results, before and after February 2022. Good news on your plots. Solo or Pool? How many tb of plots do you have?
Thanks for the info and conversation!
1 Like
Fun thread. I had not used chia plots check before. Ran it just for fun.
Anyone observed this happen before?
Proofs 36 / 30, 1.2
Proofs 42 / 30, 1.4
How is this possible? Several plots found more than 30 proofs, more typical was 31 or 32 / 30.
expected value is not the same as probability. google for Law Of Large Numbers
Yes. It shows that the millions of numbers it looks up came up at higher than expected. Very lucky plots.
1 Like
Only for that small and fixed subset of checks (those are hardcoded checks, so more or less worthless). Otherwise, as good as virtually any other plot. (Kind of debated on this forum ad nauseam.)
This is a nice video that talks about failing that plot check - https://www.youtube.com/watch?v=AojI6fixYq0
Saying that, when I tested my plots with only 5 checks, I saw few plots failing those tests. I started replotting those, but in the middle of that replotting, I checked those “bad” ones again for 30 checks, and all passed with normal scores (around 1).
hello
5 check is not good enough to test.
use for k32 30 minimum.
k33 and k34 use 100 minimum.
that is acceptable and will give you a better picture.
good luck
1 Like
a bad plot can in fact hit, but dev’s have confirm 70 is lowest you should accept. I have 5000 plots at 80 or better. my decision but it does not hurt.
good luck.
Yeah, I was looking for bad plots or duplicates, not really testing them for “quality.” So, once I saw those “bad” results, I panicked, and started replotting them. Kind of mid-way, I run on those plots checks for 30, and all ended up good, so I aborted replotting.
So, let’s say that you test a k32 plot with 30 checks, and it will give you .65 (ready to schedule to replot). Right after you do another check, but this time starting from 50 but again with 30 checks, and you get .95. Is that plot good or bad?
Interesting proposition. I looked for documentation on the “check” method and didn’t find anything describing what those outputs actually mean. I’m not a math person but what I fail to understand is how a model can report that your probability of winning something is more than 100% Is that what “Proofs x / y” is actually saying?
I read the Law of Large Numbers and understand that. But having studied a bit of probability, I know that the chances of something occurring can never be greater than 1.00. If a model is saying the plot will win 36 / 30 times, isn’t that flawed? At best it should be 30/30. No matter how many times you roll a 6-side die, you will never roll a 7. Maybe after trillions of attempts, a model might predict that 1 out of 2 times a 1 will be rolled on the first throw, but it would be flawed to say you have a 2/1 chance of rolling a 1 on the first try, right?
1 Like
I tested one plot for 30 checks once starting with 0 and another one starting with 50 (so they don’t overlap), and the first gave me 0.9, the second 1.1
So that was a “good” plot (above 0.7), but we can easily assume that it could be a 0.6 type plot, but again the second check would be at 0.8, so what gives.
The difference between the first and the second test is just the seed used to generate those checks (semi-random generation, in this case resulting in the same sequences for the same seeds). So, claiming that the first test was “better” is basically making a statement that the first seed was better, what is rather a complete BS.
Therefore, for me, that plot check is only useful to check for plot corruption (as in that video), and even trying to use it for that is kind of bad test (as I tried to outline in the previous post when it produced several bad plot results (with 5 tests), where all plots were good when more checks were done (30)).
By the way, the example you provided is just a small outcome exercise (just 6 values), and it is as you described. In this case, there is a bazillion of hashes in each plot, so the test is not to find individual hashes, but rather check whether there are hashes in some ranges, so it is possible that in a given range you may have more than one hash. However, with a small range tests it is rather difficult to say what is the distribution of those hashes (therefore we see those differences when different seeds are being used), as well as results are different when more checks / ranges are being used.
Saing all that, I am not saying that my math knowledge is better than yours, just the results we get are contradicting what is being claimed, and looking at the relevant code (e.g., what is in that video) shows how checks are generated.
1 Like
.95 is excellent! as long as its .79 for most people and .80 for me.