I recently updated my spreadsheet (since now the FRR wall has been broken a few times) to compare your bot returns with 30 day auto-FRR:
The statement on your website ("* FRR Return Difference Assumes a 70% Efficiency for time money spends lent out. This is actually MUCH higher than most users get using the FRR, but seems like a reasonable comparison") is completely wrong, here are actual differences between your retuns and mine in percent ( -(100%-my_returns/your_returns) ) since you report them on your page:
01.12.2014 -7,443460%
02.12.2014 -10,054463%
03.12.2014 -15,504663%
04.12.2014 -16,924298%
05.12.2014 -10,553673%
06.12.2014 -27,546879%
07.12.2014 -29,276524%
08.12.2014 -33,408200%
09.12.2014 -33,379851%
10.12.2014 -25,599959%
11.12.2014 -11,808594%
12.12.2014 -11,556969%
13.12.2014 -11,719647%
14.12.2014 -11,798841%
15.12.2014 -9,980899%
16.12.2014 2,099330%
17.12.2014 1,963954%
18.12.2014 1,618318%
19.12.2014 9,932038%
20.12.2014 10,659478%
21.12.2014 17,783933%
22.12.2014 19,632251%
23.12.2014 -0,274929%
24.12.2014 36,185919%
25.12.2014 47,920128%
26.12.2014 21,771841%
27.12.2014 33,091841%
28.12.2014 24,557408%
29.12.2014 42,748676%
30.12.2014 22,251068%
31.12.2014 22,583313%
01.01.2015 19,038134%
02.01.2015 16,775189%
03.01.2015 18,753871%
04.01.2015 20,981465%
05.01.2015 10,475503%
06.01.2015 -4,716642%
07.01.2015 -9,402876%
08.01.2015 2,937789%
09.01.2015 1,264892%
10.01.2015 -7,299879%
11.01.2015 4,262874%
12.01.2015 6,635660%
13.01.2015 7,094768%
14.01.2015 3,554590%
15.01.2015 12,009196%
16.01.2015 8,244013%
17.01.2015 0,757309%
18.01.2015 1,362987%
19.01.2015 0,743891%
20.01.2015 -3,301981%
21.01.2015 3,499526%
22.01.2015 -3,648495%
23.01.2015 -17,951652%
24.01.2015 -3,792983%
25.01.2015 1,869511%
26.01.2015 4,668778%
27.01.2015 1,901741%
28.01.2015 3,410024%
29.01.2015 5,961179%
30.01.2015 -0,522030%
31.01.2015 -6,395268%
01.02.2015 -4,071584%
02.02.2015 -2,634039%
03.02.2015 0,237169%
04.02.2015 -3,479075%
05.02.2015 -7,313430%
06.02.2015 -11,208395%
07.02.2015 -5,065939%
08.02.2015 3,776792%
09.02.2015 1,576461%
10.02.2015 -5,731081%
11.02.2015 -9,831616%
12.02.2015 -7,923571%
13.02.2015 4,206058%
14.02.2015 1,688207%
15.02.2015 -5,998144%
16.02.2015 -25,304583%
17.02.2015 -18,174385%
18.02.2015 -9,773438%
19.02.2015 -3,527300%
20.02.2015 -4,175197%
21.02.2015 -5,676997%
22.02.2015 -4,951486%
23.02.2015 -10,979872%
24.02.2015 -16,248698%
25.02.2015 -25,884357%
26.02.2015 -19,517822%
27.02.2015 -14,240242%
28.02.2015 -31,930358%
01.03.2015 -36,312944%
02.03.2015 -30,285468%
03.03.2015 -17,201920%
-30% for example means that I earned 30% less that day in swaps than your bot did, +30% means I earned more than your bot. While FRR seems to definitely underperform if the FRR wall is broken (like recently), in a downtrend the FRR seems to be actually the better choice (last week of December) or at least not much different from your bot (January until mid-February).
I'd recommend a less aggressive strategy when the rates are trending down (FRR falls slower than a bot undercutting it) but it seems to be great in situations like right now where FRR might be a loss. To determine the strategy, you might want to sample and store the currently lowest rate on the books as well as the current FRR rate every 10 minutes, then calculate the return of the last 24 hours with each strategy, and either place FRR bids (FRR wall existing, FRR returns of the last 24h were above lowest rate returns) or undercut the current best offer (no FRR wall, lowest rate return of the last 24h was above FRR return). Maybe with an optional setting to adjust the time window (24 hours, 72 hours, 1 hour...)