Ah, Thanksgiving. FINALLY! A day off to
spend with the family dig up some NES dupes from the collection and record some video for statistics.
For this session we selected two of each title and only attempted each ONCE, blowing in only one of each title and leaving the other alone ("stale"). That's 24 titles, 48 game paks, simple pass/fail with no reseating what-so-ever (the crux of the confirmation bias argument). A "0/0" means that both the stale and the blown in copy failed to boot first try. A "0/1" indicates that only the blown copy booted first try. A "1/0" indicates that only the stale copy booted first try. "1/1" means they both worked first try.
0/1 Castlevania
0/1 The Legend of Zelda
1/1 Super Mario Bros.
1/1 Metroid
0/0 Golf
1/1 Dragon Warrior
0/0 Gyromite
1/1 Batman
1/1 Duck Hunt
0/1 Kid Icarus
0/0 Dr. Mario
1/1 Tetris
1/1 Faxandu
0/1 Little Nemo Dream Master
1/1 Spy Hunter
0/1 Track & Field II
0/0 Mickey Mousecapade
0/1 Top Gun
0/1 Galaga
0/0 Ghost 'N Goblins
0/0 Ivan "Ironman" Stewart's Super Off-Road
0/1 Blades of Steel
0/1 Rygar
1/0 WWF Wrestlemania
6 cases where neither worked (0/0)
9 cases where only the copy blown into worked (0/1)
1 case where only the stale copy worked (1/0)
8 cases where both worked (1/1)
As you can see, if we dismiss the results where nothing changes (0/0, 1/1), the relevant results are 9 to 1 in favor of blowing. Only one title set worked without blowing and failed when blown and it was the very last game tested. With a sample size of 48 game paks, we have a clear trend developing. We may be out of dupes but we can test hundreds of random carts the same way to easily show that the trend continues as the sample size increases. Because we would be dismissing same-title comparative statistics, the results we dismissed would be relevant (some successes could potentially be from blowing, for example).
If you don't want to compare pass/fail by title, then we have 17 pass first try when blown versus 9 which passed first try without being blown. Keep in mind that this NES has had the 10NES chip disabled which GREATLY increases success for all cartridges. If blowing actually does improve chances for a good connection on all pins then this modification will make the possible improvements from blowing LESS noticeable. If there is a positive effect from blowing then disabling the 10NES lockout chip should not be needed as often for the blown cartridges, but even so we still see an 89% greater success rate for cartridges which were blown in! This is a significant difference even considering the sample size.
While there can obviously be out-right fraud, with this testing methodology there can be no accidental confirmation bias. You can attribute it to randomness (FLUKE!) but I will point out that the deviation would skew 50/50 across multiple tests with much slighter deviations so I will ask that you show a contradictory data set. I will also increase the sample set with further random testing. Go ahead and accuse me of fraudulently cleaning only one set but I will insist that none of the game paks were cleaned prior to testing and none have been cleaned for years (if ever). The only deliberate selection for which game pak went in which stack was Super Off-Road when we noticed electrical tape crammed along the edge connector. We deliberately placed it in the "blow" pile, so that it would work AGAINST confirmation bias for the outcome we expected (and observed anyway). Mickey Mousecapade was NOT a title I had possibly blown in to test recently despite wondering that in the video. When assembling sets I collected three copies and, like all titles where I had three or more copies, I deliberately used ones that were in storage as opposed to using the accessible one that could have been cleaned, blown in, tested, or played in past years. I started two piles so that titles which were blown in for recent testing would go in the blow pile in case of a residual effect. Those would be from my accessible shelf copies, but they were gradually eliminated from the set as I found third and fourth dupes or removed if I did not find a second copy. Neither Off-Road nor Mickey title sets worked anyway (all failed with and without blowing), so they did not skew the results toward my expectations.
By the way: you guys owe us a new power button! We wore it out with all our testing.

It randomly felt "crunchy" the last time we tested a ton of carts hundreds of times and it failed to power up once during this session.
Video coming soon. I started recording after my brother so I missed the first couple games. I'm waiting on his video but I didn't wait to tally the results.
Edit:
http://youtu.be/36uocRNwLnw
My brother still hasn't uploaded the complete video but I figured I'd go ahead and put up my late-start video.