Developments for the Business Planning Add-on: Bidding Analysis April 25, 2019 John D. Grace 2019 GOM3 User Conference
Bidding Models 1.0 Bid/No-Bid model estimates: Bid-Amount model From the public information we can measure characteristics of blocks with respect to the variables we think bidders associate with profitability of exploration on those blocks Companies preferences with respect to these characteristics that we measure through analysis of where they did and didn’t bid in the past. Bid/No-Bid model estimates: Probability that each open block receives ≥ 1 bid Probability that each company will bid on each open block Top 20 players modelled individually Next 20 players modelled as single “composite” company Classification of each company revised annually Bid-Amount model Forecasts most likely value of high bid for each open block Produces a single estimate for all bidders (i.e., “high bid” is estimated but not do not try to forecast who submitted the high bid for each block) Advantages of general approach Results based on objective, consistent & transparent methodologies Provides benchmarks & evaluation of outliers Forecasts with explicit measures of uncertainty Replicability for validation and sensitivity analysis 2019 GOM3 User Conference
Grouping Bidding Companies Individually Modelled Companies Company Money Exposed (Million $2017) Number of Bids Money Exposed (Million $2017) High-Activity Companies Medium-Activity Companies (cont) Shell $1,419 285 Ecopetrol $212 121 BP $777 336 Stone* $195 87 Equinor $895 189 Cobalt $179 108 Chevron $889 284 Venari $175 62 ExxonMobil $914 239 Red Willow $170 79 BHP $618 223 Total $158 120 Anadarko $443 185 Murphy $156 64 Medium-Activity Companies Repsol $138 74 LLOG $306 115 Ridgewood $124 67 Noble $237 84 Houston Energy $59 89 Hess $265 69 Low-Activity Companies Modelled Collectively Apache Davis Fieldwood Nexen Tana Arena Deep Gulf Focus Nippon W&T ATP Energy Res. Tech. Gulfslope Petrobas Walter Calypso ENI MCX Rising Nat. Res. Wild Well CL&F Enven Navitas Talos Woodside
Bid/No-bid Methodology Logistic Model Equation 𝑌 𝑖,𝑡 = 1 1+exp( − 𝛽 0,𝑡 + 𝛽 1,𝑡 𝑋 1,𝑡 +…+ 𝛽 𝐾,𝑡 𝑋 𝐾,𝑡 ) + 𝜀 𝑡 Global variables: Oil price Change in oil price Local company variables: Current co. leases Past co. leases Past co. relinquish. Past co. bids Co. AOI Local general variables: Water depth Distance to nearest: Platform Discovery Drilling Field Current, past & relinquished leases Past bids Newly available block Run by Company Top 20 companies run individually Next 20 companies run as a single “composite” co. Co. classifications changed annually 2019 GOM3 User Conference
Bid/No-Bid: Results (Sale 249) Anadarko Training Period: 2014 – 2017 (Sale 247) 2019 GOM3 User Conference
Bid/No-Bid: Results All Companies Training Period: 2009 – 2013 – Forecasting: 2014 2019 GOM3 User Conference Year: 2016 Year: 2015 Year: 2014 Year: 2017
Bid/No-Bid Accuracy 2019 GOM3 User Conference
Bid Amount Methodology Linear Model Equation 𝑌 𝑖,𝑡 = 𝛽 0,𝑡 + 𝛽 1,𝑡 𝑋 1,𝑡 +…+ 𝛽 𝐾,𝑡 𝑋 𝐾,𝑡 + 𝜀 𝑡 Global, local & company variables: Largely same as in Bid/No-bid model Analysis done in “real” dollars Y Greater emphasis on near-by & past bids: Largely same as in Bid/No-bid model Estimated collectively for all companies for each block 𝒍𝒐𝒈(𝒀)=𝟖.𝟕𝟓+𝟎.𝟎𝟎𝟕𝑿 X 2019 GOM3 User Conference
Bid Amount: Results (Sale 249) Bid Amount: Distribution of Bid Amount Training Period: 2014 – 2017 (Sale 247) 2019 GOM3 User Conference
Problem of Extreme Bids $630K Over-estimate $980K $70K Over-estimate $80K $16.7 Million Under-estimate Hess: $18.3 BOEM (MROV): $3.6 Chevron: $2.8 ESA: $1.8 Shell: $1.6
Bidding Models 2.0: New Directions Competition: Relation of number of bidders & bid amount? Integration of geoscience into models Explore relationship between of bid/no-bid & high bids Changing statistical method: more robust for outliers Revise rule on removing blocks from analysis Provide user more detail on output 2019 GOM3 User Conference
Bidding 2.0: New Module: Rejections Between 5 – 7% of high bids are rejected as “too low” Gov’t raises MROV near established production Gov’t sets >80% of frontier MROVs to min. legal bid Two modeling approaches: Forecasting the MROV on each open block Forecasting the probability that MROV = MLB Different statistical model: decision trees/random forests 2019 GOM3 User Conference
Proximity to Productivity Raises MROV Red: PUQ Light Blue: PUQ - Ring1 Dark Blue: PUQ - Ring2
Frontiers Are Highly Bimodal 2009 - 2018
Will MROV = Min. Legal Bid? 0% NO YES Block rejected last time? Block unleased > 3 years? Block next to > 1 current lease? 88% 70% Block next to lease with MROV > $11 mil.? 90% 63% 2019 GOM3 User Conference
Probability MROV = MLB
Estimated MROV & Possible Rejections