SEPTEMBER 2020

DEVELOPER ACTIVITY REPORT

Table of Contents

Introduction 3

Developer Activity to Market Cap Ratio as a Leading Indicator 4

Challenges of Analyzing Commit Data as a Proxy for Productivity 5

Correlation Between Developer Activity and growth 6

Modeling a Hypothetical 2.5 Year Fund Based on These Metrics 7

Tokens with Highest Average DMR Over Last Month 8

Other Applications of the Data 9

Delphi Data Dashboards for Institutional Members 10

Analysts
Luke Saunders Jose Macedo Piers Kicks

2
 Introduction
Developer activity correlates to future growth
As researchers covering the blockchain space, our goal is to find great projects
early. We’re fortunate to have been able to do this with REN, SNX, RUNE, NXM and Projects with market caps between $1m and $5m over the last two years
others. However, with an average of 70 new tokens listed each month, it’s difficult
for us to dive into each one at the depth required to independently gauge their
potential.
Projects with < 1x price growth
Most of the time, our discovery process flows through our pre-vetted network built
through years in the industry. However, we realize that this is a lagging rather than 0.72 Average daily active developers
a leading indicator and we became fascinated with finding leading indicators that Average daily commits
could allow us to find outstanding projects even earlier in their lifecycle. 2.39

We intuitively know that developer activity is a key fundamental indicator for the
success of blockchain projects. Its importance is already recognized for large cap Average
projects, with Electric Capital releasing their widely read developer reports.
However, we found research on small cap project developer activity to be virtually 0.99
non-existent.
4.06
This surprised us, as we understood that developer activity was even more
important for early stage projects whose success depends on building and iterating
quickly to arrive at a quality product. We also recognized that the payoff of finding Projects achieving > 10x price growth
these projects early was enormous given the explosive returns of small cap projects.
1.83
In this report we begin by exploring the difficulty of uniformly quantifying
developer activity due to the fungible nature of commits and lack of a unique id for 8.39
work, as well as our solutions to this challenge.

We also explore the correlation between developer activity and a project’s future Case study: Thorchain (166x growth)
success, particularly regarding token price performance where we see a notable
and we believe actionable correlation for projects with low market cap to developer 6.15
activity ratios. Surprisingly, almost no correlation exists for higher market cap
projects and we devote some time to discussing why this might be the case. 39.78

Lastly we introduce our new Delphi Data dashboards for exploring developer
activity, allowing our members to track developer activity for projects in real-time.

3
Sources: internal git data analysis, CoinGecko as of 28th August 2020
 Developer Activity to Market Cap Ratio as a Leading Indicator
We derive a rating from developer activity and market cap and test its correlation to future growth.

DMR rating – a comparable score How DMR score relates to market cap growth

In order to effectively explore the correlation We found that a higher DMR score correlates directly to a higher multiple return. In a period where
between developer activity, market cap and the market largely remained flat, outsized returns could be achieved from projects with low market
future growth for small projects, we need a
caps and high development activity. Notably, a strategy of buying projects with a DMR > 20000
simple metric that represents how favorable the
ratio of developer activity to market cap is for
achieved 31.05x returns, outperforming BTC by nearly 3000%.
any project, at any date. Average return when buying every time the score is over the given threshold and holding until now. Bitcoin and DMR > 0 included as a base where
DMR > 0 represents buying everything. Tokens only bought if listed within a year prior and their market cap is over $1m. Dates prior to 28th June 2020
DMR has four inputs: and projects listed earlier than 28th Feb 2018 ignored.

• Project & date Bitcoin (DCA every day)

DMR > 0
• Total commits on that date across all
repositories that the project owns, excluding 1.71x DMR > 1000

forks and any copy/pasted repositories DMR > 2000

3.71x DMR > 5000
• Unique developers who submitted the above
DMR > 8000
commits 4.96x
DMR > 15000
• Market cap on that date for that project DMR > 20000
5.74x
Commits and active developers are normalised
between 0 and the 99.5th percentile in order to 8.18x
strip outliers.
DMR is then calculated per date per project as 10.76x
follows:
23.19x
𝒏𝒐𝒓𝒎𝒂𝒍𝒊𝒔𝒆𝒅 𝒄𝒐𝒎𝒎𝒊𝒕𝒔 + 𝒏𝒐𝒓𝒎𝒂𝒍𝒊𝒔𝒆𝒅 𝒂𝒄𝒕𝒊𝒗𝒆 𝒅𝒆𝒗𝒆𝒍𝒐𝒑𝒆𝒓𝒔 ∗ 𝟎. 𝟓
𝑫𝑴𝑹 =
𝒏𝒐𝒓𝒎𝒂𝒍𝒊𝒔𝒆𝒅 𝒎𝒂𝒓𝒌𝒆𝒕 𝒄𝒂𝒑 31.05x

1x - break even point

0 5 10 15 20 25 30 35

Average return measured by market cap growth as a multiple of initial purchase

4
Sources: internal git data analysis, CoinGecko as of 28th August 2020
Challenges of Analyzing Commit Data as a Proxy for Productivity
Using commit data is challenging for a number of reasons such as fungibility, rebasing and forked projects.

Forks Branch rebasing Other challenges

Some crypto projects are forks of others. For example several A developer creates a new branch and while working, the master Cherry-picking
projects are forks of Bitcoin and others of Ethereum. To make branch moves on. In some cases a commit, or set of commits
things more complex, these projects also might have clones
ba35f0 879c2a 7e0078 can be picked from one branch and
of Electrum or other projects within their Github/Gitlab
committed into another.
organization. Master branch
In this case the author and committer dates
Bitcoin commits QTUM commits vary so by ignoring commits with differing
Feature branch committers and authors, this work would be
counted only once.
4d9a1c2 657d393 bd4008
dd0ee5 450b99 ed706c 506f13 69f859 83f9a8
The branch is rebased, which leads to the existing commits in the branch
In this example, there is no way to distinguish between getting new hashes and committer dates making them appear like new Commits are fungible
Bitcoin commits and original commits by the QTUM team. work
Not all commits are equal – one commit
The problem gets worse when new work from the Bitcoin ba35f0 879c2a 7e0078 might contain hundreds of lines of changes,
team is merged into the QTUM project by the QTUM team.
and another just single character changes.
Master branch
This is partly mitigated by analysing daily
active developer counts, though we are
dd0ee5 8f37b5 69f859 83f9a8 817bd1
Feature branch looking at including lines-of-code changed as
450b99 ed706c 506f13
df1bdd 64d0b0
another metric to weigh against.
02ab2b
This is why the QTUM project on Github appears to have 673 To avoid double counting the same piece of work, we employ two strategies:
contributors and 24.934 commits. In reality it has a fraction of
this. 1. When analyzing a repository for the first time Lines of code

While counting lines-of-code would be a

Count all commits where the author and the committer are the same
useful metric, often compiled or binary files
person, even when committer dates and author dates differ. This is
are committed which have a large LoC
okay as in the case of a rebase, the original commits are no longer
footprint which representing little actual work.
visible in the history so can’t be counted twice.
To mitigate this we manually identify when a repository is a
clone of another, then mark the parent repository in our
2. When updating stats for an existing repository
database. Our platform clones the parent repository and
does not count any commits which were made by a
When scanning a repository every day, it would be possible to count
contributor to the parent repository.
the commits as they are initially made, then again after rebasing. To
mitigate this we ignore all commits where committer and author date
differ, so that the post-rebase commits are ignored.
5
 Correlation Between Developer Activity and Growth
We discovered our hypothesis was correct: there is a notable positive correlation between developer activity
and returns in small cap crypto projects.

Relationship of DMR to ROI for projects between $1m and $10m market cap using weekly snapshots
256 • Each dot represent an individual project’s DMR and
future price growth till present for a specific week over
the last 30 months.
128
Price growth multiple to present (log scale)

• For projects with market caps lower than $10m this

shows that the correlation between developer activity
64 and returns is strong. That is, projects which a high
ratio of developer activity to market cap tend to go on
to see a high price growth.
32
• However for projects with higher market caps, the
correlation is much weaker. This might indicate a
16 lower importance of developer activity and a higher
importance of real traction for more mature projects.

8
> $10m market cap
256
4
128
64
2 32
16
8
1 4
2
1
0.5 0.5
1000 4000 16000 64000 1000 2000 4000

DMR score (linear scale)

DMR score (log scale)

6
Sources: internal git data analysis, CoinGecko as of 28th August 2020
 Modeling a Hypothetical 2 Year Fund Based on These Metrics
We back test a scenario where tokens are blindly bought whenever their DMR rating is > 10000 and sold six months later.

Projects bought Fund cash balance and NAV over the two years until present day

Name Times Average

Bought Return 4,500,000
Total Value Cash Balance Book Value
elrond 28 27.05
darwinia-network-native-token 18 17.65
4,000,000
thorchain 37 6.66
livepeer 19 5.92
fetch-ai 50 5.41
nimiq-2 135 5.16
3,500,000
bosagora 1 5.14
energi 36 5.05 overall return of 20.15x
solana 9 4.53 3,000,000
kadena 10 4.28
dusk-network 10 3.09
commerceblock-token 31 3.06 2,500,000
raven-protocol 5 2.95
origin-protocol 17 2.84
dock 89 2.82 2,000,000
loki-network 16 2.79
kleros 19 2.68
idena 2 2.35
1,500,000
ocean-protocol 28 2.29
anj 10 2.21
harmony 10 2.12
robonomics-web-services 4 1.96 1,000,000
neumark 9 1.74
genesis-vision 145 1.72
raiden-network 9 1.64 500,000
cartesi 30 1.34
mcdex 2 1.33
dev-protocol 46 1.33 0
aergo 5 1.26 9/9/18 2/9/19 7/9/19 12/9/19 5/9/20
wownero 1 1.11
keep-network 2 1.09 Fund buys Energi Energi is sold Fund buys Thorchain Fund buys Elrond
haven 1 0.99 every day for a between September between March and
datamine 1 0.88 month and November 2019 May 2020
2key 24 0.87
pillar 2 0.78
opacity 6 0.77
0xcert 39 0.76
vite 64 0.73
internxt 2 0.72 Starting with $150,000 on September 9th 2018, the fund ended on September 9th 2020 with a NAV of $3m, for a total return of 20.15x,
conceal 33 0.64 outperforming Bitcoin by 1215% in the same period. Though some result in losses, the average return per buy Is 3.49x over 6 months
banano 5 0.59
adamant-messenger 78 0.48
sharder-protocol 12 0.36 In reality we wouldn’t allow our fund to buy Adamant Messenger nearly 80 days in a row, which shows the importance of secondary
open-platform 4 0.33 layers of due diligence. Additionally a ‘many bets’ style strategy may be effective with low cap projects where the upside potential
contentos 3 0.28
appears to outweigh the losses.
dos-network 12 0.24
7
Sources: internal git data analysis, CoinGecko
 Tokens with Highest Average DMR Over Last Month
Four projects with DMR scores of over 10000 over the last month show potential for a deeper look

Importance of second filter 70,000 * Excluding projects which have been listed for more than 365 days or have a market cap of below $1m

60,000
• Our hypothesis is that, rather than blindly
following the screener, returns can be 50,000
significantly improved by adding a second

DMR
filter in the form of an additional layer of 40,000
manual analysis. In our Monthly Developer
30,000
Activity Report, we take all projects above
10,000 DMR and analyse their product- 20,000
market-fit, token economics, issuance and
team, to generate a vetted list of high DMR 10,000 Buying above this line has historically given an average return of 4.72x over 6 months
projects.
0
• The below is a sample of the kind of 7/31/20 8/3/20 8/6/20 8/9/20 8/12/20 8/15/20 8/18/20 8/21/20 8/24/20 8/27/20
analysis we’ll provide, based on the six
projects that currently have DMR scores Dev Protocol 2key CommerceBlock Robonomics Keep Network Aragon Juror
above 10,000
Keep (KEEP) Aragon Court (ANJ) Robonomics Web Services (RWS) 2key (2KEY)

Keep Network is a privacy layer for Aragon Court is Aragon's decentralized Robonomics is an open-source platform 2key is a decentralized affiliate link platform
blockchains allowing users and apps to dispute resolution service. With Aragon and decentralized network of IoT providers benefiting both sides with trustless
store data. KEEP is a work token entitling dominating the DAO space with over 1500 which can connect a robot as a service for automated affiliate payments and a
users to participate in the Random Beacon DAOs and 14/20 of the top DAOs by end-users or digital markets. RWS is the reputation system helping ensure that
and signing for tBTC. tBTC is a an app built treasury, we see the Court as an extremely subscription token where 1 RWS entitles the clicks are not faked.
which lets Bitcoin holders deposit and important step in building Aragon's holder to one transaction per second
redeem BTC on DeFi without centralized network effect. from/to their IoT device over the The project uses a plasma chain as a layer 2.
intermediaries. Robonomics Polkadot parachain. Most of the development happens in
ANJ itself is a capital asset entitling jurors to private repos so its DMR score would
Keep is close to re-launching tBTC on all fees generated by the court. This, The RWS token has a limited supply of only actually be higher than shown here if all
mainnet after earlier issues. A point of combined with the fact that Aragon is 100 tokens leading to a 6 figure price per repos were public.
concern is the high issuance with 20% of about to launch Agreements, means we token. Unfortunately its DMR score has
the total KEEP supply being distributed in may begin to see a significant increase in dropped sharply in recent days as its price However after some initial signs of traction,
a frontloaded manner over a span of 6 – 12 traction for the court. and market cap have increased by over fundamentals have dropped with
months to ETH stakers and Playing for 900% showing the importance of acting clickthroughs per day dropping to a fraction
Keeps monthly winners. quickly in identifying undervalued microcap of their 4 figure high and there are
assets. significant token unlocks happening this
year. 8

Sources: internal git data analysis, CoinGecko as of 28th August 2020

Other Applications of the Data
While using git data to identify promising new projects is our first application, the potential use cases of this data are endless

Other Use Cases

ETH2.0 Client Development Comparison

Active developers per day (filtered)

35
• In this report we focus on using
developer activity as a leading 30
indicator for low market cap
25
projects, however we expect the
data to be useful for various other
20
usecases.

• To the right here we show two 15

charts from the upcoming
Ethereum report as examples of 10
how developer activity can serve as
5
a proxy for measuring a project’s
rate of progress.
0
5/24/16 11/24/16 5/24/17 11/24/17 5/24/18 11/24/18 5/24/19 11/24/19 5/24/20
• In coming months we will explore
other ways that developer activity
chainsafe-eth cortex-eth ethereum-trinity harmony-eth2 parity-shasper pegasys-eth prysmatic-labs sigma-eth status-nimbus-eth yeeth-eth
might correlate to price: does a
drop in development activity tend
Active developers per day (filtered)

to be a leading indicator of a drop Programmable Layer 1 Development Comparison

in an asset’s price? Is a high level of 100
development activity relative to a 90
project’s competitors an indicator
that this project will gain market 80
share? 70

• Delphi Institutional Clients will 60

have access to the live data 50
dashboards, while Institutional+
clients will also be able to access 40
the raw CSV files in order to 30
conduct their own analysis on the
data. 20

10

0
5/5/16 11/5/16 5/5/17 11/5/17 5/5/18 11/5/18 5/5/19 11/5/19 5/5/20

cardano cosmos elrond eos ethereum harmony kadena near nervos-network ontology polkadot solana tezos
9
Delphi Data Dashboards for Institutional Members
Institutional members will benefit from dashboards giving access to a real-time view of the data referenced in reports

Developer data is the first of many proprietary data products Delphi will roll out
which will be guided by and complement the research. These will be made
available to our subscriber tiers as described below, with insights members having
access to reports while institutional members have access to real-time tools.

Insights Institutional Institutional+

$50/month $500/month > $2500/month

Monthly Dev Reports

Real-time Data Dashboards

CSV Exporting

The Delphi Daily

Insights Reports

Report Walkthroughs

Monthly Analyst Calls

Analyst Telegram Access

Economic Models

10
 85 Broad Street
New York, NY 10004
www.delphidigital.io

11