Hmmmm there's something about that thread that seems a bit off to me.Just something about the way the op speaks about CS make it seem like he has some vested interest.
Like this comment "reading reviews by people that only use CB I guess, the GSA name is what
sells it, and the sales thread. everyone I know who has both says that
CS is way better, there just to busy making money so don't post reviews
:P"
A new test would be great. One thing that is true is that CB "thinks" it recognises many more than it actually does. A quick glance at the log shows tons are wrong but still marked as correct.
I have both now and must say the CB seems to be more stable and also recommended by a friend who actually builds his own OCRs ...here's a part of the conversation i had with him:
[22-Nov-13 11:40:55 PM] MX: told you to get the gsa braker first works out of the box with you ubs
[22-Nov-13 11:40:55 PM] MX: gsa captcha breaker is much better
[22-Nov-13 11:40:59 PM] MX: i have both
[22-Nov-13 11:41:09 PM] MX: captcha sniper will also crash after a few days of running -
[22-Nov-13 11:41:13 PM] MX: captcha breaker you can run forever ... ... ... for some unknown reasons CS does crash from time to time and (not confirmed yet) i think it skips captchas.
I have read that thread yesterday and was a bit upset that people still belief that CS guy.
Nevertheless I downloaded that html file they included in that thread as "proof" and bingo....another attempt to make CS shine in a brighter light where it is clearly not as good as CB.
Why? Because if you have a look at the captchas (right click in browser to show details) it shows it is always a PNG image type. Thats not normal as PNG is not used for all captchas but also JPG and GIF. So they used the imorted captchas from there CS database (Mikey or anyone having access to there collected captchas) and send that to CB.
Now as only PNGs got sent and no program was active that is known to be a "ReEncoding one" (some programs always encode a captcha to png/jpg and send that to the captcha service) it though that the captchas are original and not modified/re-encoded.
And that skipped most of the captcha types as captchas defined to only get in JPG e.g. will not even used for the captcha type detection.
This is the reason for the many empty and wrong replies.
I don't feel like posting on there thread as I really can't stand this manipulations and childish behavior. But if anyone from here feels like telling the truth...
Ya, I knew there was something fishy about the whole thing also.
Well I guess that's what happens when you have competition in the industry. @Sven, would you mind if I replied in that thread with what you said here and see what that guy says about it?
LMAO. I posted a reply in the thread on BHW pretty much the same thing that @Sven said, and the OP from there already replied.
@Sven, your right about everything you said. I wonder if this guy is actually affiliated with CS in some way.
In my reply there I said:
Now the ONLY real test would be to load up 1000 websites and NOT captchas and do this test.
And here is his reply to that:
This method of testing is extremely flawed as the programs won't get the same captchas so you won't be able to directly compare them. But I might do a test like this soon in the future.
) ) ) )
Now is not the point of doing tests like this to actually test it against websites to see what program will work better?
Don't believe everything you read, we run in excess of 100k contextual link per day and on a performance and cost-effective basis nothing compares right now to CB in terms of stability and success rate, except maybe xR (though it can't be directly used a an external tool to solve captchas).
Yea it's just sad that most people belief fancy graphics and do not ask the right questions as they don't know the internals. Anyway thanks for pointing it out there.
OK one thing I have to add after reading the replies on that thread.
The option you can turn on in CB to ignore the image types has no influence if no detected program is running that reencodes images. That detection is done while scanning all actively running programs and there executable names.
So turning the option on and send it many reencoded png files from the CS database is just not working.
I'll be more than happy to test with other captcha's if anyone here can send me over some in the correct format. I also found another thread here https://forum.gsa-online.de/discussion/2264/cb-vs-cs-test-results-running-both-softwares-on-500-sample-captchas that was the same test I did, but no one questions the validity of that test. and yet when CS beats CB everyone suddenly jumps in and questions every small detail.
I am not biased towards any tool and I will use the one that yields the better results and at the moment that is CS, when its CB I will switch over and use that again.
Why do you use captchas in png format from CS guys? Do you call that a legit comparison? Even if I gave you my captchas (in poper format not reencoded) I would not trust it as it's from CB developers.
It's always a question on how well you prepare your comparison. And this is in no way any proper setup.
We have several other tests on this forum done by independent guys. And CB wins always. And I did not influence anything. Even if CB lost, I would try to see why and improve the captcha results.
Right now I don't see any reason to improve anything...since the results are better in CB than CS. Test it properly and you will see, if you are interested in this at all and no CS fan boy as everyone thinks.
@cool_dude123 Why not test them both with a seed list of a few thousand websites? That's how to do a real test to see what program will solve the most captchas.
That's too obvious and too easy, much better to pre-select which captures you want to test. Even better if you pre-select them from the DB of one of the softwares you are testing, that gives a much fairer result )
@Sven I used the captcha's from the CS guy because he was the only guy I knew that had lots of captchas. I use the tool that gives me the best results and at the moment its CS, maybe I have some really bad settings in CB? I don't know. I bet those tests you are referring to compare against old versions of CS and only use a small select number of captchas. I will happy accept captchas from you if you want to send them over and I'll do the test.
also @Sven how did you come up with your solve rates in CB? I would be interested to know as I am not getting close to that amount when testing, but then it must be because its a PNG file and for some unknown reason even though the image is exactly the same as the original it solves it differently
@Trevor_Bandura, I used to use CB then switched to CS. After switching I got a lot more submissions. that lead me to dig deeper and try to find out how big the difference was. I still don't have a complete understanding about the difference in both tools as I have been told my testing has been flawed so I am now trying to find out how to make the tests more reliable and more accurate. when I get a nice list of like 1,000+ sites I will happily post to them to check both tools (but I currently have a few important projects running)
@gooner, yeah I sat down for hours picking out 15,000+ captchas that work well in CS but not well in CB.
)
ok guys, tell me how to do the tests and supply me with the captchas or the site lists and I will even livestream it all for you
@cool_dude123 - I'm not suggested you hand picked captchas that work well in CS. But i am suggesting that the creator of CS who sent you the captchas did EXACTLY that, it would be in his best interests to do that.
And i'm not against you or CS but if someone ran a test and got all the captchas directly from Sven, i would be wary of that result too.
Let SER scrape random sites with random captchas and run your test, that would be fair and that would be a result that matters to all us SER users. That's all i'm saying.
@JamPackedSpam CB shit yes? Just try to add a bit more arguments or leave it.
@cool_dude123 Im sorry but I will not give you my captcha db. The reason is simple. We all know Mikey steals every new captcha type someone shared online with our program. And as I think you are either Mikey or his close friend/fan, I will not make it that easy for him. So sorry, no captcha from me. But why not turning on the "save captchas" in CB? Don't do it in CS as I guess it saves the reencoded one, but in CS you can get the original captchas and work on that.
About solve rates, I don't understand what you are referring to here. If a captcha is skipped it is not taken for the % calculation if thats what you meant.
>@gooner, yeah I sat down for hours picking out 15,000+ captchas that work well in CS but not well in CB.
Yes you did, or well Mikey did that for you. As he trained his image filters on the captchas he got. Thats normal of course as he has to work with something at least. And it would be the same if I give you mine. The way I define the image filters is based on the test captchas I have.
>ok guys, tell me how to do the tests and supply me with the captchas or the site lists and I will even livestream it all for you
See what we all mean? You ask us how to do tests. Do you call this in any way professional? You should take care of this yourself before creating such false conclusions.
I never said I was professional, just a guy who did a test
"such false conclusions."
I can only conclude from what I have tested and that is what I have concluded
"If a captcha is skipped it is not taken for the % calculation if thats what you meant."
oh, so your solve rate is only from the captcha's that CB attempts. so if a cpatcha is too hard and only attempts 1 of out 100 and gets it right then it would be a 100% solve rate and not 1%?
>oh, so your solve rate is only from the captcha's that CB attempts. so if a cpatcha is too hard and only attempts 1 of out 100 and gets it right then it would be a 100% solve rate and not 1%?
Should we really discuss this? You get the calculation all wrong. Please do your home work first.
@cool_dude123 You said that you switched to CS over CB because you were getting more submissions, well that's the test that everyone needs to see to show with no dought what software is better.
Exactly what I've been saying all along. Not to sure why that's so hard to understand. After all, these two software's are supposed to help us make more submissions without having the need for a paid third party captcha solver.
@Trevor_Bandura - That isn't a good way to test because there are too many random variables during submissions. Even if you import the exact same URL list and post using CB, then post using CS, results are not going to be accurate because both software didn't post using the same captcha. One might have received an easier version of the captcha than the other.
The only real way to test would be to take the exact same captchas and run them through both software. By exact same captchas I don't mean the trained captchas that were handed off by the developer of CS as .PNG files to skew the results.
Best thing to do would be to download your own set of captchas for the types you want to test. For example, you could download say, 500 wordpress, mediawiki, guestbooks, etc. etc. captchas then run those through both software.
The only problem is it takes a long time to download so many captchas and then rename each one to have the correct answer. That's probably why most people don't test properly.
Of course you could take them from the CB folder but those are all recognized so again it might skew the results as well. Best way would to get a fresh downloaded list of all the captcha types you want to test so nothing is "fishy".
This is why the developer of CS linked to that "test" in his sales thread and told all his customers if they want to try it and test it for themselves to contact him so he can send them captchas to test with ... hmmm ... wonder why that is?
Comments
Like this comment "reading reviews by people that only use CB I guess, the GSA name is what sells it, and the sales thread. everyone I know who has both says that CS is way better, there just to busy making money so don't post reviews :P"
Smells like bullsh....
One thing that is true is that CB "thinks" it recognises many more than it actually does.
A quick glance at the log shows tons are wrong but still marked as correct.
I have read that thread yesterday and was a bit upset that people still belief that CS guy.
Nevertheless I downloaded that html file they included in that thread as "proof" and bingo....another attempt to make CS shine in a brighter light where it is clearly not as good as CB.
Why? Because if you have a look at the captchas (right click in browser to show details) it shows it is always a PNG image type. Thats not normal as PNG is not used for all captchas but also JPG and GIF. So they used the imorted captchas from there CS database (Mikey or anyone having access to there collected captchas) and send that to CB.
Now as only PNGs got sent and no program was active that is known to be a "ReEncoding one" (some programs always encode a captcha to png/jpg and send that to the captcha service) it though that the captchas are original and not modified/re-encoded.
And that skipped most of the captcha types as captchas defined to only get in JPG e.g. will not even used for the captcha type detection.
This is the reason for the many empty and wrong replies.
I don't feel like posting on there thread as I really can't stand this manipulations and childish behavior. But if anyone from here feels like telling the truth...
OK one thing I have to add after reading the replies on that thread.
The option you can turn on in CB to ignore the image types has no influence if no detected program is running that reencodes images. That detection is done while scanning all actively running programs and there executable names.
So turning the option on and send it many reencoded png files from the CS database is just not working.
Why do you use captchas in png format from CS guys? Do you call that a legit comparison? Even if I gave you my captchas (in poper format not reencoded) I would not trust it as it's from CB developers.
It's always a question on how well you prepare your comparison. And this is in no way any proper setup.
We have several other tests on this forum done by independent guys. And CB wins always. And I did not influence anything. Even if CB lost, I would try to see why and improve the captcha results.
Right now I don't see any reason to improve anything...since the results are better in CB than CS. Test it properly and you will see, if you are interested in this at all and no CS fan boy as everyone thinks.
@Sven I used the captcha's from the CS guy because he was the only guy I knew that had lots of captchas. I use the tool that gives me the best results and at the moment its CS, maybe I have some really bad settings in CB? I don't know. I bet those tests you are referring to compare against old versions of CS and only use a small select number of captchas. I will happy accept captchas from you if you want to send them over and I'll do the test.
also @Sven how did you come up with your solve rates in CB? I would be interested to know as I am not getting close to that amount when testing, but then it must be because its a PNG file and for some unknown reason even though the image is exactly the same as the original it solves it differently
@Trevor_Bandura, I used to use CB then switched to CS. After switching I got a lot more submissions. that lead me to dig deeper and try to find out how big the difference was. I still don't have a complete understanding about the difference in both tools as I have been told my testing has been flawed so I am now trying to find out how to make the tests more reliable and more accurate. when I get a nice list of like 1,000+ sites I will happily post to them to check both tools (but I currently have a few important projects running)
@gooner, yeah I sat down for hours picking out 15,000+ captchas that work well in CS but not well in CB.
)
ok guys, tell me how to do the tests and supply me with the captchas or the site lists and I will even livestream it all for you
And i'm not against you or CS but if someone ran a test and got all the captchas directly from Sven, i would be wary of that result too.
Let SER scrape random sites with random captchas and run your test, that would be fair and that would be a result that matters to all us SER users. That's all i'm saying.
@JamPackedSpam CB shit yes? Just try to add a bit more arguments or leave it.
@cool_dude123 Im sorry but I will not give you my captcha db. The reason is simple. We all know Mikey steals every new captcha type someone shared online with our program. And as I think you are either Mikey or his close friend/fan, I will not make it that easy for him. So sorry, no captcha from me. But why not turning on the "save captchas" in CB? Don't do it in CS as I guess it saves the reencoded one, but in CS you can get the original captchas and work on that.
About solve rates, I don't understand what you are referring to here. If a captcha is skipped it is not taken for the % calculation if thats what you meant.
>@gooner, yeah I sat down for hours picking out 15,000+ captchas that work well in CS but not well in CB.
Yes you did, or well Mikey did that for you. As he trained his image filters on the captchas he got. Thats normal of course as he has to work with something at least. And it would be the same if I give you mine. The way I define the image filters is based on the test captchas I have.
>ok guys, tell me how to do the tests and supply me with the captchas or the site lists and I will even livestream it all for you
See what we all mean? You ask us how to do tests. Do you call this in any way professional? You should take care of this yourself before creating such false conclusions.
"Do you call this in any way professional?"
I never said I was professional, just a guy who did a test
"such false conclusions."
I can only conclude from what I have tested and that is what I have concluded
"If a captcha is skipped it is not taken for the % calculation if thats what you meant."
oh, so your solve rate is only from the captcha's that CB attempts. so if a cpatcha is too hard and only attempts 1 of out 100 and gets it right then it would be a 100% solve rate and not 1%?
>oh, so your solve rate is only from the captcha's that CB attempts. so if a cpatcha is too hard and only attempts 1 of out 100 and gets it right then it would be a 100% solve rate and not 1%?
Should we really discuss this? You get the calculation all wrong. Please do your home work first.