Tactics improvement in 2 graphs

I've blogged previously about my long chess break from February to September and how it appears that break did me good.  I was messing around with some more Chess Tempo output and decided to graph my non-duplicate accuracy and average solve time by problem rating.

There definitely seems to be about a 3% accuracy increase and 5 second speed increase.

Maybe there is a case for taking regular long breaks with chess playing/training?  Recently I came up with the idea of breaking my chess into seasons, just like sports.

Off season: Do nothing (maybe light annotated games reading and some Bain level tactics)
Pre-season: Lots of new hard training (Yusupov, non-duplicate CT problems), lots of reading and LTC training games at ICC.
In-season: Lots of OTB, "maintenance" training only (just keeping up with repetitions?)

So how long should each phase be?

Still in the saddle

Nothing too earth shattering or important here, I just wanted to get in the habit of updating the blog regularly.  I decided I wanted to dive into my Chess Tempo stats a little more since I completed an entire calendar month of tactics on 10/31.

I ended up doing exactly 2,700 tactics for the month of October with 552 non-duplicates.  That represents the second most amount of tactics I've ever done in a month (December '14 with 2,743) and the third most amount of non-duplicates in a month.  Of course, it was much easier to do non-duplicates when I was first getting started.  Every day I have a several custom sets I go through.  I always do at least 20 problems from each set.  If the amount of scheduled problems is more than 20, I just do the scheduled problems.  If it's less than 20, I do new problems until I've done 20 in that set.

Instead of looking at my performance stats by custom set, I decided to break them down individually by the motif I was training.  This means I get the motif stats across any custom sets.  Since there can be a lot of overlap in tags, it's not uncommon for me to get lots of pins in my mate in 2 set, or distraction motifs in my "Decoys" set.

I broke the stats into 3 phases:

1. Beginning (any problem I did up until 12-31-2014)
2. Break (any problem I did up until 2-28-2015)
3. Return (any problem I've done since 9-1-2015)

The percentages represent the % of problems I got correct in 15 seconds or less with that motif.  Since some of the motifs have such a small sample size, I decided not to include all of them.

CORE (1200-1600 rating)

Distraction	40.5%	38.6%	32.3%
Fork	42.7%	41.0%	42.9%
Pin	48.5%	45.5%	48.9%

DECOY (1200-1600 rating)

Attraction	53.3%	52.3%	54.8%
Clearance	48.7%	43.2%	49.4%
Coercion	42.0%	40.2%	44.7%

With the exception of the distraction motif, they all follow the same pattern.  

1. Initial good stats at a time when I'm in the groove for tactical training.
2. Significant dropoff right before I take a break from heavy training.
3. Noticeable increase over my best performance even after a long break.

I also chose a motif that I've never specifically trained before, but consider it be important.

Discovered Attack

32.5%

29.3%

33.3%

It follows the same pattern as the other motifs.

What does it all mean?  Who knows.  I just hope I don't get burnt out again since these numbers seem promising.