I decided to try to keep track and record my flight recall stats and post my findings to be able to track performance and improvement. Here are some things I learned from yesterdays training session:
40 recalls made, 30 treats, 115g, several other tricks, motivation 4/5, almond pieces used as recall treat
Overall recall stats:
32/40 1st attempt recall 80%
7/40 2nd attempt recall 17.5%
1/40 missed recall on 2nd attempt 2.5%
Early part of training session with fr1:
5/8 1st attempt recall 62.5%
3/8 2nd attempt recall 37.5%
Mid session recalls at vr3:
10/15 1st attempt recall 67%
5/15 2nd attempt recall 33%
Mid session recalls at fr1:
6/8 1st attempt 75%
1/8 2nd attempt 12.5%
1/8 missed attempt 12.5%
End of session recalls at fr1:
8/9 1st attempt 89%
1/9 2nd attempt 11%
I have to say she was pretty motivated this session and did better than expected. With the exception of one flight where she botched the landing and landed somewhere else, she always came within 2 calls and came on the first call 80% of the time. Her hunger, motivation, and focus where better than average and I focused primarily on recall during this training session which probably made results better than typical.
A very interesting discovery I made during this training session is that 1st attempt recall
improved in the course of the training session rather than dwindle as I would have predicted. There was clearly some learning going on there. Also, it would appear that the frequency/ratio of reinforcement had
no effect on recall success ratio. It seems that the improvement pattern had more to do with the number of recalls prior rather than the frequency of reinforcement. The implication is that more or more frequent rewarding does not necessarily improve recall response. With VR3 the parrot could do more recalls for the same number of treats so the value is a longer training session while satisfying the same hunger.
I stopped at 40 recalls because I felt like she had reached her peak and did not want to reach a point where motivation would decline. It is interesting to note that immediacy of recall continued to improve all the way up to the 40th recall rather than drop off from getting bored or tired. I will try to continue making these recordings and see what I find out.