Nearly all bitcoin miners have been transferring their rewards within 36 hours of obtaining them, researchers at New York University have found.

Luqin Wang and Yong Liu's paper, an analysis on the "evolution" of pool mining, found that less than 1% of miners left their rewards untouched in 2012 and 2013.

Those who did transact with their rewards did so within a week in 2012, and within 36 hours the following year. The researchers did not perform the analysis for 2014.

Miners' hasty transfers in recent years contrasts sharply with the behaviour of early miners. In 2009, 66% of miners left their rewards untouched. Those who did make a transaction did so about four months after obtaining the reward.

The following year the share of "frozen miners" fell to 20%. By contrast, in 2011 just 2% of miners didn't perform a transaction with their mined rewards.

The paper suggests that one reason early miners haven't moved their bitcoins is because they have lost access to those wallets, locking them out permanently. These early miners probably got involved in bitcoin not for financial gain, but because it was a "fun technology" that could be played with casually, the authors write.

From 2009 to 2013, the bitcoin price exploded, from several cents a coin to $1,128 at its height in December 2013. This price surge was accompanied by a rise in the number of miners joining the bitcoin network. As the number of miners grew, miners went from going "solo" to joining pools, in order to share their computational power while maintaining their chances of capturing a reward.

The authors analysed miners' profitability as the bitcoin price rose over the years. They then looked at two pieces of mining hardware – a Radeon graphics card and a Butterfly Labs ASIC miner – and compared their performance to the bitcoin price and network difficulty in 2011 and 2013.

The paper will be presented at the Passive and Active Measurements Conference, which discusses network measurement and analysis techniques, in New York on 19th March.

Wang and Liu found that the early miner with the graphics card, if paying the average United States electricity price, broke even on their investment in just under two years. The owner of the ASIC miner, however, could have broken even in less than a month, if the hardware was purchased in July 2013. Profits would have continued to accumulate well into the following March, where the researchers' analysis stops.

The pair also dig into the mechanics of the Discus Fish mining pool, which consistently had the biggest chunk of the network hashrate in the second half of last year, a title it recently ceded to Antpool.

Discus Fish, which also goes by F2Pool, saw a surge in the number of miners in its pool at the end of 2013, the authors find. The figure rose from a few hundred miners in October to 5,000 just five months later.

An analysis of the distribution of rewards within the Discus Fish pool is able to estimate the hashrate of the top 10% of miners within the pool compared to the pool averages. The top 10% of miners in Discus Fish have a hashrate that is about a thousand times greater than the pool average, the paper found.

Since the pool employs a 'pay per share' model, miners are rewarded in direct proportion to the number of shares, or computational power, they contribute to a pool. Therefore, the authors conclude, the most powerful miners in a pool also reap most of the rewards.

An independent analyst of mining pools, who goes by the pseudonym Organ of Corti, said the paper had implemented a method to quantify the average hashrate of individual miners within a pool for the first time.

"They implemented a method to determine a miner's average hashrate by monitoring F2Pool's output transactions," the analyst said. "It's not a new idea, but it's the first time I've seen it done."