I looked over generate_nn_social.py and wonder why is for average block time estimation a depth equal to the blocks count till the planned hardfork used? Why not just take average for some rather long past period, like the whole year?
It seems to me that the longer time span the more precise estimations would be.
(btw is this a new algo or we use it every year?)

I looked over generate_nn_social.py and wonder why is for average block time estimation a depth equal to the blocks count till the planned hardfork used? Why not just take average for some rather long past period, like the whole year? It seems to me that the longer time span the more precise estimations would be. (btw is this a new algo or we use it every year?)

I can do it over a longer period, though thought that more recent blocks may have a greater weight for recent conditions. I'll do a few more time frames to see how it compares.

Its a "new" algo - in the past I believe a simple assumed 1 block per minute was used, but last year due to slow block production it was inaccurate. Since then production has stabilised (we added some things in last years HF to help).

I was actually surprised to see the average over the last 6 weeks was pretty close to 60 sec, only 2 extra seconds.

I looked over generate_nn_social.py and wonder why is for average block time estimation a depth equal to the blocks count till the planned hardfork used? Why not just take average for some rather long past period, like the whole year? It seems to me that the longer time span the more precise estimations would be. (btw is this a new algo or we use it every year?)

I can do it over a longer period, though thought that more recent blocks may have a greater weight for recent conditions. I'll do a few more time frames to see how it compares.

Maybe good to compare with the same period before the past hardfork

Double checked that there were no errors in any of the keys submitted via:

#!/usr/bin/env python3

import json
import base58
import bitcoin
from bitcoin.wallet import P2PKHBitcoinAddress
from bitcoin.core import x
from bitcoin.core import CoreMainParams


with open("socials.json") as key_list:
    socials = json.load(key_list)

class CoinParams(CoreMainParams):
    BASE58_PREFIXES = {'PUBKEY_ADDR': 60,
                       'SCRIPT_ADDR': 85,
                       'SECRET_KEY': 188}
bitcoin.params = CoinParams

for op in socials:
	for region in socials[op]['regions']:
		for party in socials[op]['regions'][region]:
			key = socials[op]['regions'][region][party]
			assert str(P2PKHBitcoinAddress.from_pubkey(x(key['pubkey']))) == key['KMD address']

class CoinParams(CoreMainParams):
    BASE58_PREFIXES = {'PUBKEY_ADDR': 48,
                       'SCRIPT_ADDR': 50,
                       'SECRET_KEY': 176}
bitcoin.params = CoinParams

for op in socials:
	for region in socials[op]['regions']:
		for party in socials[op]['regions'][region]:
			key = socials[op]['regions'][region][party]
			assert str(P2PKHBitcoinAddress.from_pubkey(x(key['pubkey']))) == key['LTC address']

Checked blocktimes over the last 360 days alongside each 30 day slice within that timeframe.
https://0bin.net/paste/EpHWcGsa#Pehpn6jG4jSEZPPKFkMsAUtoie+s7v7R8Enhs6B0wwg

HF estimated block range: 3484507 - 3485925
HF Estimated block average: 3485590

Open to any other methods or suggestions.

Considering the fact that block generation is more stable this season,
I think the proposed estimation using the recent blocks should be okay