A Minor Winter

“It was the best of times, it was the worst of times,
it was the age of wisdom, it was the age of foolishness,
it was the epoch of belief, it was the epoch of incredulity,
it was the season of Light, it was the season of Darkness,
it was the spring of hope, it was the winter of despair,
we had everything before us, we had nothing before us…”

While the bear market falls over the past few months, people assumed a massive wave of investment and miners leaving crypto currencies domain. From data, however, we noticed that the trend is not that severe and miners found out the way to survive with profits still, by mining empty blocks. We analyzed why this behavior is more profitable by statistics, and identified the major players who are following this practice.

Winter is coming. How are miners doing?

For the past 6 months, it’s no secret that everyone’s had their eyes fixated on the falling price of Ether. We decided to take a look behind the curtain to better understand the tangible impact the events of recent months have had on miners and mining pools. Let’s dive into the the mines and see what the past few months have looked like underground.

Number of Blocks Stays Stable

A high-level view of the network for the past 6 months indicates that the number of blocks stays at the same level, and so does the number of transactions processed per day, at around 600,000.

Fig. 1 — Block Throughput Of The Network Stays Stable(Q3-Q4)

This doesn’t mean that mining activities volume stays the same though, as the Ethereum difficulty adjustment algorithm keeps tuning the difficulty of mining to keep the mining time within a stable range of 10 to 19 seconds. Thus as the number of blocks is measured over time, the block rate always keeps steady.

Mining Activities Decrease

Meanwhile, what can tell us about the fall in mining activities is the drop in difficulty and hashrate — though it doesn’t necessarily show a decrease in number of miners, it certainly reflects the drop in the number of mining machines. Here by the “miner”, we mean the address that mined blocks on chain, instead of real identity behind it. The addresses in most cases are mining pools, with hundreds of individuals registered in them contributing hashing power.

Difficulty measures the complexity of the task that miners need to solve to mine the block. Hashrate totals the computing power of miners in the network. When the mining activities decreases, hashpower decreases, therefore the time of solving problems of the same difficulty as before would grow. The unit used to measure difficulty is TH (terahashes — trillions of hashes, or trillions of attempted solutions). Thus, the network hashrate totals the number of attempts all miners make, and is measured in trillions of hashes per second (TH/s).

Fig. 2 — Average Difficulty Of Mined Blocks & Hashrate Of Network per Day Correlates and Dropped

Above we can see that along with the fluctuation of hashrate across network, the difficulty also changes over time, showing two major drops in September and November. Similar positive correlation showed in the scatter plot below.

Fig. 3 — Average Difficulty & Hash Rate Of Each Day

An extra scatter plot of each block’s difficulty and hash rate shows a clear positive correlation between them as we explained theoretically earlier. An interesting finding is that they fall into 2 clusters with the diverging showing around difficulty of 3200 TH. — We believe this results from the reaction lag of miners towards the sharp adjustment of network’s difficulty, especially when it goes up — from Fig 2, we see that both of the steep slopes (in June & September) of difficulty happened around 3200 TH, where the orange hash rate value failed to adjust closely but with a time lag to react and keep up. Whenever there’s a sharp decrease the miners can just stop contributing hash power when it’s enough, but when difficulty is going up it’s hard to estimate where it will end up and they might over- (in September) or under-(in June) estimate it. There might be more interpretations and assumptions around this pattern to be investigated furthermore. (Let us know your guess!)

Number of Miners (Mining Pools) Stays Stable

The chart below shows that the average number of distinct miner addresses hasn’t changed lately. This may indicate that miners responded to market conditions not by quitting, but by scaling down their operations. And that’s a good sign, it may show that their confidence in the network hasn’t wavered.

Fig. 4 — Number Of Miners And Their Payout Addresses By Month

This was expected — after all, they’re still playing the same game and they’re still in it for the profits, so by design mining activity is supposed to go down when profitability rates go down. Mining rigs use a lot of electricity, and this adds to the costs a miner has. Furthermore, after ETH goes below a certain threshold, the math goes like this: If I, as a miner, pay X amount of $ a month for electricity to mine 1 ETH, and the market price of that same 1 ETH is less than X — then why mine it, and not just buy it?

Number of Mining Individuals (behind Mining Pools) Drops

Another good way to quantify a miner operation’s eventual downscaling is by the number of payments it makes to it’s hashers — the individual miners in the pool. In the previous chart we see a clear and smooth drop for that count.

The table below shows some metrics for the top 5 mining pools by hash power — recorded in the first and second half of the year, and compared.

Fig. 5 — Summary of Top 5 Mining Pools by Hashing Power in 2018 (Q1 & Q2 vs. Q3 & Q4)

One thing in common for them is the significant decreases in the number of accounts that they make payouts to, the highest one is f2Pool.

Fig. 6 — Number Of Payout Accounts For Ethermine In Q3 & Q4

Let’s take the case of Ethermine, the largest mining pool on this date, the chart above shows a constant decrease in the daily number of payouts made. The spikes we see are regular increases in the number of addresses Ethermine makes payouts to — once every week, there is an increase of around 20–25% in the number of outgoing transactions. Their average payout however, stayed at around 0.15–0.16 ETH, including those days.

Fig. 7 — Snippet For Ethermine Account From Ethstats — Number Of Outgoing vs. Incoming Txs

Are mining pools mining empty blocks?

Another metric that highlights the changes in the mining ecosystem is the recent increase in the number of empty blocks mined. There are a couple of explanations for why miners might adopt this strategy intentionally, and a number of effects to which it gives birth.

Fig. 8 — Number of Empty Blocks In The Past 6 Months

Can mining empty blocks be more profitable? Statistically Yes.

From the miners perspective, we can see that the average total reward that miners collect in a certain period is made up of 3 different components: the block reward (~89%), the fee reward (2%), and the occasional uncle reward (~9%). To calculate a miner’s average reward per block, the best indicator would be the average of block and uncle rewards (including their respective average fees) — weighted by the miner’s respective uncle rate (since it differs from miner to miner, the network average will do for this example).

It comes out at 3.03 ETH — or around $275 for the current ETH-USD rate (as the date of Dec 11, 2018). This would be how much a miner earns per block. Also, we compared the average producing time of blocks and found that for blocks with less transactions, the average producing time is shorter. For those blocks with or less than 10 transactions, the average block producing time is 9.8 only, while the average time for all blocks since June ( block 5710964 to 6710964) is 14.5 seconds.

Meanwhile the 11,741 empty blocks mined in this period took 13.2 seconds on average.

Besides, as the time is shorter, empty blocks secure miners with a lower possibility of being verified as uncles, but more likely take the advantage of time and become the main chain.

Therefore mining empty blocks rather than normal ones can provide the miners a higher efficiency regarding profit (rewards/time is higher).

Who is mining empty blocks?

With another closer look, we noted that across the past 6 months, there are 103 miners mining those 11,741 empty blocks. The major players among them are as follows:

Fig. 9 — Miners Who Mined Empty Blocks Over Past 6 Months

The top 5 addresses are:
F2pool_2 (1st — 0x829bd824b016326a401d083b33d092293333a830),
Etherdig (2nd — 0x8d35067233605bef6069191ae0922d134ff80d48),
Ethermine (3rd — 0xea674fdde714fd979de3edf0f56aa9716b898ec8),
Dwarfpool (4th — 0x2a65aca4d5fc5b5c859090a6c34d164135398226)
Nanopool (5th — 0x52bc44d5378309ee2abf1539bf71de1b7d7be3b5)

Not surprisingly, 3 of these (F2Pool_2, Ethermine and Nanopool) are also found in the current top 5 by hashpower. From the following chart, we can see that, F2Pool_2 in particular showed a great increase in empty block production lately. As we recall from the summary of the top 5 from earlier, they were also the miner that came off the worst during this latter half of 2018.

Fig. 10 — Changes In Empty Block Output Per Top 5 Miners — Overall vs. Past 3 Months

Diving deeper, and using the fact that more empty blocks reflect directly in the decrease of average gas usage we can plot the gas usage in F2Pool_2 blocks against the ETH — USD prices chart.

Fig. 11 — Average gas usage %in F2Pool_2’s blocks in 2018

The chart above brings into focus a couple of steep, and obvious changes in the transaction inclusion strategy of F2Pool_2: in late March, late June, July, and September. Now, let’s compare these to ETH — USD prices.

Fig.12 — ETH /USD Prices In 2018

By following the biggest downward movements, we notice the same pattern: dips in March, July, and September — so it would be apparent that F2Pool_2 adopted a very price sensitive transaction inclusion strategy.

However — the second largest empty block miner, Etherdig, doesn’t use this strategy to adapt to price movements — it simply uses it as it’s main model. They mined an average of 86% empty blocks throughout 2018 (compared to F2Pool_2’s 5.5%). While, the “role model” among the top 5 mining pool by hashing power — SparkPool — didn’t show up as top player in mining empty blocks at all.